初创公司尽调
尽调报告 AI / generative video software (China) Series C private company at roughly unicorn scale after March 2026 financing 2026-07-03

AIsphere

有真实产品和融资动能,但这个独角兽标的披露仍不足

AIsphere 的产品、规模与融资证据足以让它保持可投资性;但当前独角兽级价格在 ARR 质量、毛利韧性和证券条款上仍缺文件支撑,还不足以给出买入结论。

封面要素

最新轮次 01
300 USD M Series C (Mar 2026) [CO015, CV001]
主要估值信号 02
1000 USD M~ (directional unicorn-level mark; exact post-money undisclosed) [CO037, CV002]
据报 ARR 下限 03
40 USD M+ at end-2025 / early-2026 public reporting [CO018, CV003]
据报用户规模 04
100 M+ global users [CO019, CV003]
据报 MAU 05
16 M+ monthly active users [CO019, CV003]
成立时间 06
2023-04-07 [CO002]

公司概况

AIsphere(北京爱诗科技有限公司)是一家位于北京的生成式视频创业公司,2023 年 4 月由创始人兼 CEO Wang Changhu 创立, 联合创始人 Xie Xuzhang 也在 2026 年融资报道中公开露面。旗舰产品 PixVerse 覆盖文本生成视频、图像生成视频、创作者模板、 移动端分发、企业风格 API,并通过 PixVerse R1 延伸到实时世界模型。公开融资节奏很快:2025 年完成大额 Series A 轮, 随后是 Alibaba 领投的 Series B 轮,再到 2026 年 3 月 CDH 领投的 $300 million Series C 轮。因此,公司的公开故事是真实且已起规模的, 但收入质量、毛利结构、治理和安全条款仍缺少足够披露。

官网
aishiai.com
成立时间
2023-04-07
创始人
Wang Changhu, Xie Xuzhang
创立地点
Beijing, China
总部
Beijing, China
产品
PixVerse 将自研 AI 视频基础模型与 Web/移动端创作工具、创作者模板、团队和画布工作流、API 访问,以及实时世界模型分支 R1 组合在一起, 试图把公司从短片生成推进到交互式视听世界。
客户
面向消费者创作者和专业创作者、社交视频用户、营销人员和 SMB 内容团队、授权 IP 活动,以及需要可扩展视频生成能力的企业/API 或合作伙伴工作流。
商业模式
商业模式最可能是消费者订阅和点数、高阶套餐升级、企业/API 打包,以及合作伙伴 / 授权工作流的组合,而不是单一消费者应用业务;具体渠道结构仍未披露。
阶段
Series C private company / unicorn-level headline valuation frame
融资情况
公开披露的融资包括截至 2025 年 3 月合计 >CNY400M 的 Series A 轮、2025 年 9 月 Alibaba 领投的 $60M Series B 轮, 以及 2026 年 3 月 CDH 领投的 $300M Series C 轮;最新一轮被广泛描述为中国 AI 视频赛道的创纪录融资。
[CO001, CO002, CO003, CO005, CO006, CO007, CO015, CO018]

执行摘要

主要优势

  • 产品表面真实存在,覆盖消费者 app、网页、模板、API 与实时世界模型研发,而不是单一 demo 模型。
  • 对一家成立两年的中国 AI 视频初创来说,公开融资动能异常强,最终落到由 CDH 领投的 $300M Series C。
  • 据报道的规模标记(>100M 用户、>16M MAU、>$40M ARR 下限)如果经验证,就有意义,足以支持继续尽调而不是直接放弃。
  • PixVerse 出现在独立 benchmark 和消费者 app 生态里,说明产品故事不只是公司自己的营销。

主要风险

  • 公开证据仍无法对齐 March 2026 轮融资的精确投后估值、优先股堆叠或证券质量。
  • >$40M ARR 数字与公司口径有关且未经审计,收入究竟是耐久的企业 / API 支出,还是更脆弱的创作者需求,仍未说清。
  • AI 视频竞争仍激烈,OpenAI、Runway、Kling、Wan、Hailuo、Jimeng/Seedance 与 Pika 都在场,定价权和锁定效应受限。
  • 即便用户增长保持强劲,中国监管、标识、版权与先进算力依赖风险也可能压低变现和毛利质量。

未决问题

  • 按产品表面拆分的经审计 ARR 桥、渠道组合、毛利率与单位经济性仍未披露。
  • March 2026 条款清单、投后股权结构表、清算瀑布与附函未公开。
  • 公开来源未披露付费客户数、留存 / NRR 或头部客户集中度。
  • 董事会构成、合规审计材料与详细信任安全落地,在已抓取记录中仍不完整。

目录

Chapter 01

01公司概况

1.1 身份、创立与产品范围

公开资料可以追溯到 AIsphere 即北京爱诗科技有限公司,这是一家 2023 年 4 月在北京成立的创业公司。公司官网称其在打造全球领先的 AI 视频生成大模型和应用,PixVerse 则是面向消费者创作和企业 API 工作流的外层产品。公开发布报道显示,PixVerse 在 2024 年初进入海外市场; 官方产品界面如今覆盖 Web、移动端、模板、API 和世界模型研究。这一组合很关键:AIsphere 并不是把自己包装成一次性演示应用, 而是在把自己定位成垂直一体化的视频模型公司,具备自研基础模型研发、分发界面和可变现的创作工具。抓取到的证据还显示,公司在持续维护消费者、 博客、应用下载和企业界面,这提高了一个判断的置信度:产品足迹已经在运行,而不是停留在愿景。[CO001, CO002, CO003, CO004, CO005, CO008]

快照 KPI 表
指标数值 / 状态日期置信度缺口
法定 / 公开名称Beijing Aishi Technology Co., Ltd. / AIsphere 公司2026-07-03
成立时间2023-04-072023-04-07
总部中国北京2026-03-06
当前阶段C 轮私营创业公司,处于独角兽门槛 / 约 $1B 名义估值2026-03-12抓取的公开来源支持独角兽级估值方向,但没有对齐准确的投后估值数字。
最新披露融资轮CDH Investments 领投的 $300M C 轮2026-03-12
上一轮主要融资Alibaba 领投的 $60M B 轮2025-09-24
报道 ARR2025 年底 >$40M2026-03-12ARR 来自媒体报道,未经审计,也非公司申报。
报道用户基数全球 >100M 用户2026-03-12用户数在公司相关和媒体相关报道中反复出现,但未经独立审计。
报道 MAU>16M2026-03-12月活用户出现在一条 Yicai 来源链中,抓取来源集没有独立佐证。
核心产品PixVerse 消费级 + 企业级 AI 视频平台;R1 实时世界模型;V6 电影感模型2026-03-18
公开治理可见度抓取的公开记录没有出现完整董事会名单、股权结构表或优先权栈披露。

各行混合了公司主张、媒体报道和观察到的产品界面。空值表示抓取的公开记录不足以精确核验,无法做成干净 KPI 条。

[CO001, CO002, CO003, CO009, CO012, CO015]
FO002: 公司快照逻辑

AIsphere 把自研视频模型研发、消费者分发、企业 API、战略云支持和仍待补齐的披露层连在一起。

[CO004, CO005, CO024, CO025, CO030, CO031]

1.2 领导层、创始人市场匹配与治理可见度

公开领导层记录主要围绕 Wang Changhu,次要围绕联合创始人 Xie Xuzhang。Wang 曾在 Microsoft Research Asia 和 ByteDance 任职, 让 AIsphere 的创始人履历与大规模视频和计算机视觉系统直接对上,这很可能有助于招聘和投资人信任。Xie 在 2026 年融资报道中作为公司战略、 全球扩张和世界模型雄心的公开发声者出现。缺失的信息同样重要:抓取到的公开记录没有披露完整董事会名单、确切所有权结构或优先股堆叠细节。 因此,治理和经济控制权的透明度低于融资标题暗示的水平。尽调上,没有公开董事会名单几乎和有强创始人一样重要,因为资本密集型 AI 公司进入后期轮次后, 控制权可能变化很快。[CO006, CO007, CO036, CO040]

领导层与创始人表
人物职务背景创始人-市场匹配度或覆盖关键人依赖
Wang Changhu创始人兼 CEO前 Microsoft Research Asia 研究员,前 ByteDance 视觉技术负责人与计算机视觉、大规模视频系统和 AI 人才招聘高度匹配
Xie Xuzhang联合创始人2026 年报道中公开谈及融资目标扩大、市场扩张和世界模型方向提供战略和市场端支持,但公开履历比 Wang 更薄

抓取的公开记录清楚识别了 Wang Changhu,并多次引用 Xie Xuzhang,但没有披露完整高管梯队或董事会名单。

[CO006, CO007, CO036]

1.3 融资历史、阶段与投资者联盟

AIsphere 的公开融资路径推进很快:截至 2025 年 3 月完成合计 >CNY400 million 的 Series A 轮,2025 年 9 月完成 Alibaba 领投的 $60 million Series B 轮,2026 年 3 月完成 CDH Investments 领投的 $300 million Series C 轮。投资者组合横跨财务投资人、 战略互联网 / 云支持方、国资关联基金、娱乐集团和东南亚金融机构。公开报道一致称 Series C 是中国 AI 视频行业迄今最大单笔融资。 这在方向上支撑用户给出的“Series C / 独角兽”框架,但抓取到的公开记录仍未充分披露准确投后估值和优先权条款。[CO009, CO012, CO013, CO015, CO016, CO017]

利益相关方或投资者图谱
利益相关方角色控制权或经济重要性尽调问题
CDH InvestmentsC 轮领投方锚定 2026 年大额扩张融资和估值叙事确认持股比例、董事会权利和清算优先权。
Alibaba / Alibaba CloudB 轮领投方和云合作伙伴同时提供资本和基础设施 / 合规支持确认排他性、云支出承诺和战略控制权。
Fortune Capital早期机构支持方参与出资首个披露的规模化融资轮梳理早期投资者是否在 C 轮中保留按比例跟投权。
Shenzhen Capital Group / Beijing AI Fund / E-Town Capital 等投资方国资相关资本方释放政策和生态支持信号,但可能塑造商业化预期澄清任何政策、地点或产业合作义务。
Ruyi Holdings / 37 Interactive娱乐行业投资方对媒体 / 内容分发关系可能有价值验证财务投资是否转化为分发或授权渠道。
UOB Venture Management / Lion X Fund 等投资方海外财务投资者支撑全球扩张叙事,并为 C 轮加入非内地资本判断它们增加的是渠道价值,还是只有资产负债表资本。
创始人 / 管理层运营控制公开记录识别了 Wang 和 Xie,但没有完整股权结构表获取股权结构表、董事会席位、归属安排和创始人稀释历史。

本表列出抓取融资报道中可见的主要公开具名资本方和战略利益相关方。控制权和持股比例仍是私有信息。

[CO009, CO012, CO013, CO015, CO016, CO017]
FO001: 公司里程碑时间线

AIsphere 从 2023 年创立,推进到 2025-2026 年的快速融资、产品迭代和平台扩张,同时竞争也在加剧。

[CO002, CO008, CO009, CO012, CO015, CO020]

1.4 增长信号、产品里程碑与平台演进

对一家成立两年的创业公司来说,AIsphere 的公开叙事里程碑密度异常高。到 2025 年 3 月,Yicai 报道其用户超过 40 million, 月活跃用户超过 15 million。到 2025 年底和 2026 年初,公司关联和媒体关联报道把这一叙事推高到超过 100 million 用户、 超过 16 million 月活用户,以及超过 $40 million 的 ARR。产品上,AIsphere 于 2026 年 1 月发布实时世界模型 R1, 2026 年 3 月发布更可控的电影化模型 V6;官方更新页还显示,生产平台正在扩展 CLI、团队和企业层。大趋势很清楚: AIsphere 正在以很快速度同时扩大产品范围和分发速度。[CO010, CO011, CO014, CO018, CO019, CO020]

里程碑表
日期事件类型金额 / 估值 / 状态参与方含义
2023-04-07Beijing Aishi Technology Co., Ltd. 成立创立公司在北京成立创始人 Wang ChanghuWang 离开大型平台岗位后不久,创建了一家专注 AI 视频的创业公司。
2024-01-01PixVerse 面向海外用户上线产品消费级 AI 视频产品进入市场AIsphere / PixVerse 团队给公司带来早期全球消费者切入口。
2024-03-18报道 A 轮超过 RMB100M融资SCMP 披露等值 US$14M / >RMB100MFortune Capital 领投轮尽管存在竞争质疑,早期投资者兴趣仍得到验证。
2025-03-06A 轮累计融资超过 CNY400M融资Eminence 领投 A5 轮;40M+ 用户;15M+ MAUEminence Ventures;Lighthouse Capital显示早期阶段能够持续获得资本,并保持消费者采用动量。
2025-09-24B 轮以 $60M 完成融资$60MAlibaba、Fortune Capital、Shenzhen Capital、Beijing AI Fund、Giant Network 与 Antler把头部云 / 互联网战略投资者带入公司。
2025-12-18Alibaba Cloud 全栈合作宣布合作基础设施、模型服务、产品、生态和商业合作AIsphere 和 Alibaba Cloud降低部分基础设施不确定性,同时提高合作伙伴依赖。
2026-01-14PixVerse R1 上线产品1080p 实时世界模型AIsphere / PixVerse将故事从文生视频工具扩展到交互式世界模型。
2026-03-12C 轮完成融资$300M;公开报道中中国 AI 视频最大单轮融资;独角兽级估值叙事CDH Investments + 20+ 家机构将 AIsphere 推入中国融资最多的 AI 视频创业公司梯队。
2026-03-18V6 上线,平台叙事拓宽产品电影感控制 + CLI + 生产工作流重点PixVerse 产品团队释放从新奇模板转向专业 / 生产用途的信号。
2026-03-18KrASIA 强调逆势创立和竞争压力负面据报道订阅收入覆盖成本,但竞争仍然激烈KrASIA / 36Kr 转述加入第一章的主要警示:规模化仍必须击败体量大得多的竞争对手。

这是第一章唯一的正式时间线。抓取的公开记录能把事件锚定到月份、但无法锚定到具体日期时,使用月份级日期。

[CO002, CO008, CO009, CO012, CO015, CO020]
FO003: 快照 KPI

公开 KPI 图景显示其消费端和融资势头异常快,但经审计披露有限,形成抵消。

[CO002, CO003, CO015, CO018, CO019, CO037]

1.5 反向信号与披露限制

谨慎信号主要来自战略和披露,而不是法律处罚。KrASIA 对 AIsphere 崛起的复盘强调,公司创立时,许多投资人怀疑独立视频模型创业公司能否在 OpenAI 和中国互联网巨头旁边活下来。同一篇文章称订阅收入现在已覆盖成本,但这仍是创始人口径的里程碑,不是经审计财务披露。更广泛地看, ARR、用户总数和行业领先地位等头部指标仍集中在公司关联或单链条媒体报道中。治理透明度也偏弱:没有清晰公开的董事会名单、股权结构表或准确 Series C 投后估值。这并不否定业务;它意味着第一章的置信度应建立在真实产品和真实融资之上,同时继续带着明确证据缺口。[CO032, CO033, CO034, CO036, CO037, CO038]

1.6 图表

Chapter 02

02市场分析

2.1 市场边界:生成式视频 AI 是工作流层,不是整个 AI 经济

本章最重要的纪律,是先定义边界,再谈规模。这里的窄口径市场不是全部生成式 AI,也不是全部媒体软件。TBRC 将视频创作中的生成式 AI 定义为: 用生成模型创建、编辑或增强视频内容;TBRC 和 Research and Markets 都按部署、应用和终端用户切分该类别。这些定义覆盖文本生成视频、 图像生成视频、编辑和增强、协作,以及云端或本地部署交付。它们并不能把上游 GPU 训练支出、通用超大规模云厂商 AI 收入,或非 AI 制作软件算作视频创作厂商的直接市场收入。 PixVerse、Runway、Pika、Vidu、Kling、Jimeng 和 Wan 的官方界面也显示,竞争集合如今横跨消费者创作工具、工作流平台和 API 产品。 这对 AIsphere 很关键,因为公司处在 AI 经济中一个真实但更窄的楔子里:创意视频生成以及紧邻它的工作流,而不是整个基础模型栈。[CM001, CM002, CM003, CM004, CM005, CM006]

市场定义表
细分 / 类别纳入支出排除支出买方 / 付款方与 AIsphere 的相关性
消费级创作应用创作者购买的文生视频、图生视频、模板、编辑、特效、订阅和积分泛娱乐支出、非 AI 编辑工具,以及无关社交平台广告收入个人创作者或专业消费者付款方相关,因为 PixVerse 仍在争夺创作者心智和自下而上的付费转化
SMB / 营销人员工作流营销活动视频生成、广告素材、社交内容、轻量工作流自动化与视频生成或生产输出无关的广义营销云支出营销经理或 SMB 业主相关,因为视频 AI 可以由增长预算买单,而不只依赖创作者订阅
媒体 / IP 生产协作生产、品牌模板、授权内容工作流、资产管理和分发工具与 AI 视频工作流无关的传统工作室资本开支、无关后期制作和权利成本工作室运营或媒体制作预算负责人相关,因为可控性和品牌工作流可以支撑更高价值、类似企业的预算
开发者 / API 层API 调用、合作伙伴计划、工作流集成、工具构建者使用,以及嵌入式实时视频服务视频工作流之外的通用云推理支出,或非视频开发者工具平台产品负责人或工程预算相关,因为 AIsphere 营销的是 API 和实时视频基础设施,而不只是终端用户创作
企业 / 受治理部署团队方案、治理、私有或受控部署、采购主导合同和合规工具没有视频 AI 用例的广义数字化转型预算IT、采购或业务单元负责人相关,因为企业买方更看重控制、治理和工作流可靠性,而不是一次性短片的新奇感
相邻但排除的上游层不直接计入视频应用厂商直接市场收入基础模型训练资本开支、超大规模云厂商 GPU 支出、通用公有云 AI 收入和无关 AI 软件算力和平台所有者作为投入成本和议价能力很重要,但不是 AIsphere 的干净直接 TAM

本表定义本章可计入的市场边界。它纳入已变现的视频生成工作流,排除上游 AI 基础设施和无关软件支出。

[CM001, CM002, CM003, CM004, CM005, CM006]

2.2 需要多组规模镜头,因为公开数字衡量的不是同一件事

公开市场规模只有在保留多组镜头、而不是强行压成一个 TAM 标题时,才有用。最窄的镜头来自专门的生成式视频报告:TBRC 称市场在 2025 年达到 $0.39 billion,2026 年增至 $0.47 billion,2030 年达到 $0.98 billion;Research and Markets 使用同样的 $0.47 billion 2026 起点和同样的 $0.98 billion 2030 终点。相对更大的生成式 AI 复合体,这只是一个小市场;Fortune Business Insights 估计后者 2025 年为 $103.58 billion、2026 年为 $161 billion,到 2034 年超过 $1.26 trillion。这些数字并不矛盾,更多是范围错配: 一个衡量窄应用层,另一个衡量巨大的跨模态技术栈。第三个镜头不是收入,而是采用率。a16z 称 AI 视频对短片已变得相当可靠,并指出视频公司进入消费者 AI 榜单前列;Artificial Analysis 则显示,质量已足够收敛、价格已足够分散,能支撑试用和买方比价。尽调上的正确答案是一组框架,而不是伪造一个单一 TAM。[CM007, CM008, CM009, CM010, CM011, CM012]

TAM / SAM / SOM 或规模测算视角表
发布方年份 / 时间范围地域数值增长 / 采用信号实际衡量对象局限
The Business Research Company2025 当前全球$0.39B市场在 2026 年达到 $0.47B,2030 年达到 $0.98B专门用于视频创作的生成式 AI 收入池仅是狭义品类视角;不是 AIsphere 特定份额
Research and Markets 报告2026 当前全球$0.47B预计 2030 年达到 $0.98B,CAGR 为 20.4%专门用于视频创作的生成式 AI 市场另一个狭义视角,但仍不是公司特定 SAM
Fortune Business Insights2025 当前全球$103.58B更广义市场在 2026 年增长至 $161B,2034 年达到 $1,260.15B跨多种工作负载的跨模态生成式 AI 市场远宽于视频创作,因此不应作为直接视频 TAM 使用
a16z Top 100 Gen AI Consumer Apps 榜单2025 采用代理指标全球消费者网页 / 移动端基于排名的代理指标视频产品进入头部 AI 消费者榜单,Hailuo / Kling 月访问量超过 Sora需求和使用代理指标,而不是收入规模不显示付费转化或细分结构
Artificial Analysis2026 基准测试代理指标全球模型基准测试界面引用示例中 Elo 为 1069–1096PixVerse V6 在狭窄质量带内竞争,但标价低于 Veo 3.1买方试验的性能和价格代理指标基准测试分数不等于市场份额或收入
PixVerse 官方定价2026 变现代理指标全球自助平台$1 = 5 条 V6 入门视频基于使用量的积分显示创作者进入门槛低自下而上变现和开发者 API 经济性公司自写定价不披露实际净收入或留存
Runway 定价2026 变现代理指标全球自助 + 企业$12 / $28 / $76,加企业自定义积分分层打包支持创作者、专业用户和企业扩张可比厂商定价和打包视角竞争对手定价仍不披露行业合同规模或流失

这些行有意保留不同分析单位:狭义市场报告、广义生成式 AI 背景、消费者采用代理指标和变现代理指标。它们是互补视角,不是可以相加的 TAM 组成部分。

[CM007, CM008, CM009, CM010, CM011, CM012]
FM001: 市场规模测算视角

多层视角把宽口径 AI 背景、窄口径视频市场规模,以及变现 / 采用切口分开。

本图刻意混合直接市场规模、采用和变现切口,因为已审阅公开来源没有单独给出干净的 AIsphere 专属 SAM / SOM。

[CM009, CM010, CM012, CM013, CM015, CM035]
FM002: 市场估算区间

已发布美元区间保留了窄口径视频创作与更宽生成式 AI 技术栈之间的范围差异。

各行呈现的是不同但相关的市场量区间。应把它们理解为范围和时间上的分散,而不是可相加的组件。

[CM007, CM008, CM010, CM012]

2.3 买方、用户和付款方分散在五个经济属性不同的细分市场

买方地图比随口说“创作者应用”更碎。个人创作者是真实用户;Google Play 和 PixVerse 博客也说明,模板、爆款特效和移动端工作流可以带来自下而上的采用。 但 TBRC 也点名营销和社交媒体用例,Vidu 也明确卖给广告和叙事工作流,因此营销人员和 SMB 是另一类预算方。媒体和 IP 持有者构成第三个细分市场, 因为品牌内容、授权模板和协同制作流程需要超过一次性消费者提示词的工作流控制。开发者和 API 团队又是单独一类:PixVerse 的 R1 合作伙伴计划面向工作室、 平台开发者和工具构建者,Runway 则销售实时 Characters API。最后,企业和受治理约束的团队购买逻辑不同于普通创作者;他们关心控制、 部署模式、团队访问和合同打包。实践中,本章必须区分用户、买方和付款方:终端创作者也许很喜欢产品,但经济决策通常在营销负责人、工作室运营负责人、 平台产品经理或企业采购职能手里。[CM019, CM020, CM021, CM022, CM023, CM024]

细分市场 / 买方图谱
细分买方用户付款方工作流预算负责人采用触发因素
创作者 / 专业消费者个人创作者个人创作者自助订阅或积分模板使用、社交短片、个人叙事创作者钱包或副业收入预算出片快、手机端方便、起步成本低
营销人员 / 中小企业营销负责人或中小企业主营销人员、社媒经理或自由职业者营销活动预算广告、产品推广、短视频社交内容增长或营销负责人需要用更少时间和人手产出更多视频素材
娱乐 / 媒体 / IP工作室运营或品牌内容负责人剪辑师、制片人、授权方制作预算品牌模板、协同制作、资产管理制作或内容运营负责人需要可控性、协作和品牌安全复用
开发者 / API 团队产品或平台工程负责人开发者和工具构建者平台或研发预算把实时生成或视频智能体嵌入另一款产品平台 GM 或工程总监不训练自有模型,也需要差异化视频能力
企业 / 受治理工作流IT、采购或业务单元发起人跨职能团队或内部创作者软件合同 / 转型预算受治理工作流、团队工作区、自动化和合规CIO、COO 或 BU 负责人需要治理、工作流集成和生产可靠性

买方、用户和付款方在这个市场很快分离。一旦涉及工作流治理、品牌内容或 API 嵌入,最终创作者往往不再是经济决策者。

[CM019, CM020, CM021, CM022, CM023, CM024]
FM003: 买方 / 客群地图

从随手试用走向受治理平台使用的路径,显示预算归属在哪里变化。

真实买方可以跳过步骤或并行推进,但本图捕捉的是从创作者试用到受治理部署的常见预算演进。

[CM019, CM022, CM024, CM025, CM027, CM029]

2.4 增长驱动正在把类别从短片生成推向更广的制作工作流

最强的增长驱动很清楚,也有相当充分的佐证。TBRC 明确把社交媒体和数字平台扩张列为需求驱动,两家市场研究发布方也指出自动化编辑、 个性化营销、协作和云工作流是顺风。官方产品界面解释了原因。PixVerse 现在在营销一个世界引擎,带有长时域生成、实时 1080p 和可用于生产的 API; Runway 推出通用世界模型和交互式视频角色;Pika 推出智能体和工作流自动化。这些不只是更漂亮的短片生成器,而是在尝试把视频 AI 推进可重复工作流、 交互式体验和可扩张的软件预算。定价设计也帮助采用。PixVerse 暴露低金额、基于点数的创作经济模型,Runway 则提供免费入口、专业消费者订阅和企业定制打包。 这个组合支撑经典的落地再扩张路径:创作者可以低成本试用,团队可以把工作流正式化,企业则在输出足够好且运行足够可靠后谈判更大部署。[CM030, CM031, CM032, CM033, CM034, CM035]

增长驱动与约束表
驱动因素 / 约束方向阶段影响尽调问题
社交和数字平台扩张正向当前扩大对快速、可重复短视频创作的需求量化哪些渠道真正把随手使用转成付费账号或持续企业需求。
自动剪辑、协作和个性化正向当前价值从新奇短片转向可复购的工作流软件检验相比纯生成质量,工作流功能能否提高留存和付费意愿。
世界模型、智能体和实时 API正向新兴在创作者订阅之外,打开新的开发者和互动媒体预算询问当前收入中已有多少来自 API 或制作平台使用。
低摩擦自助定价正向当前支撑从创作者切入、再扩展到团队的采用路径衡量付费转化、ARPU,以及后来升级到团队或企业计划的用户占比。
算力成本和先进芯片获取负向当前会压住利润率,并拖慢中国相关厂商迭代按产品线审查云 / GPU 承诺、每分钟模型服务成本和硬件依赖。
中国的标识、治理和备案要求负向当前把信任与合规变成持续运营成本和上线摩擦索取备案状态、审核流程和合规审查平均周期。
平台连续性和可靠性风险负向当前如果头部厂商可能改动或停掉产品,买方会犹豫核查 SLA 条款、导出选项、数据可迁移性和 API 版本承诺。
替代品拥挤和价格竞争负向当前即使需求快速增长,竞争也会压缩价值获取按细分市场梳理相较 Runway、Pika、Vidu、Kling 等替代品的赢单原因。

市场确有增长顺风,但每个正向因素都带着运营限定。在这个品类里,采用速度往往跑得比持久收入获取更快。

[CM030, CM031, CM032, CM033, CM034, CM035]
FM004: 采用漏斗或价值链地图

指数化漏斗显示,公开证据暗示从首次试用到规模化部署之间,摩擦在哪里累积。

数值是相对指数点,以创作者试用 = 100。它们不是实测转化率,只是编码公开证据中约束堆叠的位置。

[CM037, CM038, CM039, CM040, CM041, CM042]

2.5 约束:算力、监管、信任和竞争都会限制单一厂商可捕获的价值

约束栈和增长栈同样重要。信任和监管上,中国 2023 年生成式 AI 措施适用于向公众生成文本、图像、音频和视频的服务,2025 年标识规则则要求标记 AI 生成的在线内容。因此,上市速度不只取决于模型质量,也受备案、治理、审核和标识义务约束。成本和供给上,BIS 表示,即使实体位于中国境外, 向中国关联实体出口先进计算仍需许可证;外部法律评论也确认,该要求在 2026 年仍有效。因此,对中国关联厂商而言,算力可得性仍是真瓶颈。 可靠性和平台连续性也没有完全解决。a16z 称 AI 视频只是最近才对短片变得相当可靠,OpenAI 停止 Sora 则说明,即便类别领导者也可能突然改变产品方向。 最后,竞争很激烈:官网和基准测试显示可信替代品范围很宽,这意味着并非每一次使用激增都会转化为 AIsphere 的持久收入或利润率捕获。[CM038, CM039, CM040, CM041, CM042, CM043]

2.6 图表

Chapter 03

03竞争格局

3.1 格局:直接同业、替代品、内部自建和潜在进入者

AIsphere 的竞争集合异常宽,因为买方可以用多种方式完成同一项任务。最窄的直接同业桶是面向创作者的 AI 视频生成:Runway、OpenAI Sora、 Kling、Vidu、Hailuo、Wan、Jimeng/Seedance 和 Pika 都在营销文本生成视频、图像生成视频、音频或可控电影化工作流的某种组合。 当创作者、工作室或专业消费者想要打包的视频生成产品,而不是通用聊天机器人时,这些就是最干净的替代品。基准测试来源进一步显示, 中国实验室和应用如今主导了相当一部分质量前沿:Artificial Analysis 目前在多个排行榜上把 ByteDance Seedance 排在 Kling、Vidu、Wan 和 PixVerse 之前,Hailuo、Grok、LTX 和 Veo 也在不同价位持续增加选择。这意味着 PixVerse 不是在一个可以自我命名的小众类别里竞争, 而是在一个密集、快速迭代的全球赛场中竞争。 一旦纳入替代品,市场边界还会更宽。内部自建是可信选项,因为基准测试场中已经有 LTX-2.x 和 Wan 变体等开放权重模型,成熟团队可以选择自托管, 而不是购买订阅产品。模型路由平台也是另一类替代品:Runway 现在把 Kling、Seedance 和 Veo 等第三方模型打包进自己的付费计划, Pika 的 MCP 和智能体定位也公开承诺接入“所有模型”。最后,潜在进入者不需要发布独立视频应用也能产生影响。OpenAI 和 Moonshot 这样的通用实验室可以把开发者拉进更广的智能体或助手生态, 操作系统级或分发端既有玩家则可以只靠成为创作发生的默认界面取胜。对 AIsphere 来说,真正的问题不是“还有谁做 AI 视频”,而是“谁拥有 AI 视频周围的用户、 工作流和路由层”。[CP013, CP014, CP015, CP020, CP021, CP022]

竞品画像表
竞争对手 / 替代方案类别规模或战略信号目标客群差异化限制
AIsphere / PixVerse直接同业PixVerse 把自己定位为具备 API 和全球创作者触达能力、可服务企业的 AI 媒体平台创作者、专业消费者、团队和 API 构建者迭代快、价格定位强,并连接中国与全球市场不只要证明模型质量,还要证明留存和工作流所有权
Runway全球直接同业 / 聚合器$315M Series E 轮,估值 $5.3B;世界模型平台野心创作者、工作室、代理商、企业团队多模型打包、品牌化工具和企业 GTM基准优势仍可被路由到多家模型提供商
OpenAI Sora全球直接同业 / 潜在进入者曾以世界模拟产品预览,随后在 2026 年下线独立体验已在 OpenAI 生态内的开发者和创作者OpenAI 品牌、安全工具和相邻平台触达独立产品持久性看起来弱于标题叙事
Kling中国直接同业Kling 3.0 和 3.0 Omni 在 AA 排行榜接近头部专业消费者和高端创作者工作流电影感控制强、长篇分镜、原生音频公开价格点相较 PixVerse 和 Vidu 偏高
Vidu中国直接同业AA 排名质量,加上参考视频工作流定位快速周转的社交、广告和叙事创作者参考视频工作流和有竞争力的基准价格抓取到的公开治理材料相对薄
Hailuo中国直接同业a16z 将 Hailuo 列为面向全球输出的中国视频产品消费级创作者品牌能见度和中国牵引的迭代速度抓取到的落地页几乎不披露价格或合规细节
Wan中国直接同业 / 自建替代多个模型变体出现在 AA 基准中,并进入开放式替代对比混用闭源和自托管流程的开发者与创作者基准能见度强,同时系列周边有开放式替代压力公开落地页信息薄;保留来源看不清企业打包
Jimeng / Seedance中国直接同业 / 潜在进入者Seedance 在保留的 AA 排行榜领先;Jimeng 露出 ByteDance 创作者工作流基因中国创作者和 ByteDance 邻近用户排行榜领先质量,加上社区和中文提示适配ByteDance 可通过更大生态承接价值,而不必依赖独立厂商
Pika全球直接同业 / 路由层以智能体和 MCP 叙事承诺接入“所有模型”需要特效和 AI 智能体工作流的创作者可聚合模型,而非押注单一前沿模型抓取到的公开价格信息薄,定位仍在演进
内部自建 / 开放权重栈替代方案LTX 开放权重模型和进入基准的开放式替代品已进入买方比较集工作室、代理商和技术团队路由成本最低、控制力最高需要运维人才,打磨度可能落后高端应用
通用型进入者(Moonshot、Veo、Grok)潜在进入者 / 相邻替代品Moonshot 估值 $20B;Veo 和 Grok 在 AA 排名中接近前沿需要广义 AI 工作空间、而不只是视频工具的开发者可把视频与代码、搜索、推理或平台默认项打包视频可能只是众多功能之一,挤压纯视频厂商利润率

本表刻意混合直接应用、替代方案和相邻进入者,因为买方可用打包应用、路由平台或自托管栈完成同一任务。

[CP001, CP002, CP006, CP007, CP008, CP009]
FP001: 竞争定位图

以有序分值展示主要保留替代方案在开放性 / 可路由性与分发 / 信任力之间的位置。

分数是有证据支持的有序判断,来自保留的定价页、产品页、基准测试摘要和监管文件,而非供应商披露指标。

[CP020, CP021, CP031, CP033, CP035, CP040]

3.2 能力与定价:质量已经拥挤,打包方式正在变成武器

能力质量已经不足以单独拉开差距。Artificial Analysis 在宽模型场中比较质量、速度和价格,显示排行榜头部非常拥挤:Seedance 目前同时领先文本生成视频和图像生成视频, Kling、Vidu、Wan、Grok、Veo 和 PixVerse 都处在相对紧密的性能区间。这压缩了 PixVerse 拥有独特优势能力栈的说法。即使 PixVerse 表现不错, 买方也可以在同一个独立记分板上,把它直接同多个相近替代品比较,这减少了不透明价值定价的空间。AIsphere 官方材料仍能讲出可信产品故事—— 面向企业的 API、近实时生成和激进价格定位——但独立证据表明,这个类别已经变得容易比较、也容易切换。 定价和打包进一步强化了这种压力。PixVerse 公开文档称,在 Starter pack 下,$1 可生成五段 V6 720p、5 秒、无音频视频;Runway 则发布免费、 Standard、Pro 和 Max 层级,提供固定月度点数和越来越宽的模型访问。Artificial Analysis 还给出 API 层直接同业可比项:Vidu Q3 Pro 标价 $9.60 per minute,Seedance 2.0 为 $9.07,Wan 2.7 为 $16.90,多个 Kling 变体显著高于 PixVerse 自己声称的费率。 低端则有开放权重 LTX 模型,说明开发者如果接受较弱质量,可以购买或自托管便宜得多的视频生成。结果是杠铃型市场:高端品牌应用在打磨程度、 社区和工作流控制上竞争,开放或聚合栈则限制任何单一视频模型可维持的利润率。这有利于类别增长,却不利于持久的独立护城河。[CP003, CP012, CP014, CP015, CP016, CP017]

功能 / 能力矩阵
能力PixVerseRunwaySoraKlingViduJimeng / SeedancePika自建 / 开放权重
文生视频质量信号强,但产品路径不稳保留排行榜中最强
图生视频和可控性保留来源显示为中随所选模型变化
原生音频或音频联动创作保留表层资料未知可见视频和图像创作;抓取页面未强调音频可变 / 取决于模型
长篇一致性 / 分镜控制可变且更吃工程
API / 企业打包最强保留抓取集中较弱UnknownUnknownUnknown买方能自运维时较强
消费者社区 / 品牌触点品牌强,独立产品持久性弱在中国创作者语境中较强
模型路由 / 多归属支持最强
公开信任 / 合规文档弱-中买方自管

单元格是基于保留官方页面、基准摘要和法律 / 监管来源作出的有证据支撑的定性判断。“弱”或“未知”表示抓取集未证明更强主张,并不代表该竞争对手一定缺少该能力。

[CP003, CP004, CP005, CP007, CP008, CP011]
价格 / 打包比较
提供方 / 替代方案公开价格或单位信号打包模式包含能力未知项 / 含义
PixVerse$1 = 5 个 V6 720p 5s 无音频视频;企业会员 $1,500 至 $6,000 / 月积分加会员和企业层级短视频生成、API、团队 / 商业包官方价格清晰,但实际企业成交价仍未知
Runway免费层级;Standard $12、Pro $28、Max $76,按年计费每月积分加企业销售Runway 模型加 Kling、Seedance、Veo 等第三方模型路由层掌握订阅,打包能力强
Kling 3.0AA 基准视角:1080p Pro 为 $20.16 / min;720p Standard 为 $15.12 / min基准比较中的模型级 API 定价高端视频生成、Omni 变体、原生音频高端质量对应明显更高的公开单位经济
Vidu Q3 ProAA 基准视角:$9.60 / min基准比较中的模型级 API 定价文本、图像和参考视频工作流质量接近 PixVerse,同时看起来比 Kling 更有价格竞争力
Seedance 2.0 720pAA 基准视角:$9.07 / min基准比较中的模型级 API 定价品类领先的文生视频和图生视频质量如果最佳基准质量也具备价格竞争力,其他厂商的利润空间都会被压缩
Wan 2.7AA 基准视角:$16.90 / min模型级基准价格;公开落地页信息薄有竞争力的图像 / 视频质量,以及系列周边开放式替代压力基准存在感真实,但公开打包透明度有限
Pika保留抓取的首页没有清晰公开价目卡应用、实验、智能体和 MCP 入口特效、智能体工作流和多模型接入路由便利性可能比透明单位定价更重要
内部自建 / LTX-2.3 FastAA 基准视角:$2.40 / min开放权重或自托管路线成本更低的自管视频生成保留选项中最便宜的一档,会限制高端应用在商品化用例上的收费

本表比较来源集中留存的最佳公开经济指标;并非每个竞争对手都披露清晰订阅或 API 价目卡,所以这里有意混用基准 $/min 和官方积分定价,并逐项标注。

[CP003, CP014, CP015, CP016, CP030, CP031]
FP002: 功能广度 / 能力地图

就绪度视角,合并基准强度、定价清晰度、合规契合度和路由灵活性。

强、中等、弱等标签,是基于留存来源得出的证据支撑综合评分,不是供应商提供的指标。

[CP012, CP014, CP015, CP017, CP025, CP026]

3.3 分发、信任与切换成本:控制平面正在移到模型之上

分发和信任是非对称性最重要的地方。Runway 不再只是模型供应商;它在建设世界模型平台,并已经用定价把自己定位成多模型网关。Pika 通过智能体和 MCP 做类似事情;a16z Top 100 报告则认为,许多持久的消费者赢家要么路由到第三方模型,要么不靠拥有最佳前沿模型也能成功。这对 AIsphere 具有战略意义: 即使 PixVerse 在质量上跟得上,价值也可能向上迁移到路由层,或横向迁移到已经拥有用户关系的平台。同一份报告称,Hailuo 和 Kling 等中国视频产品已经走向全球; 同时应用商店正在打击山寨界面。这有利于拥有真实品牌和社区的产品,但也意味着新的消费者领导者不一定要拥有独特底层模型。 信任和监管姿态又加了一道筛选。面向中国的厂商已经处在 2023 年生成式 AI 服务暂行办法和 2025 年标识制度之下,后者要求对 AI 生成内容做可见和基于元数据的标识。 这些规则带来标识、日志、导出 / 下载控制和平台审核上的运行负担。面向美国和欧盟的企业买方则有另一套镜头:NIST AI RMF、EU AI Act 和 Article 53 式 GPAI 义务,都会把厂商推向文档、版权处理、训练数据摘要和可识别生成内容。换句话说,产品质量让厂商进入候选名单,治理流程则可能决定谁能通过采购。 公开记录在这里好坏参半:OpenAI 和 Runway 详细讨论安全或产品限制,Hailuo 和 Wan 在抓取界面上公开的治理细节少得多。如果 AIsphere 想更深入销售给企业或受监管的创意工作流, 信任姿态必须像模型演示一样清晰可读。[CP020, CP021, CP023, CP024, CP025, CP026]

FP003: 护城河 / 准备度 KPI

一组紧凑信号,说明竞争耐久性为何已不只取决于模型质量。

KPI 数值混合了基准指标、公开规则生效日期和产品状态事实,因为目标是综合判断竞争准备度,而不是做单一数字仪表盘。

[CP006, CP014, CP015, CP024, CP029, CP035]

3.4 护城河耐久性:反向证据指向商品化,除非 PixVerse 拿住工作流和合规

竞争反向证据已经足够强,商品化应被视为基准风险,而不是尾部风险。OpenAI 的 Sora 是最清晰的警示:公司先预告世界模拟叙事, 后来推出带有已承认物理弱点和安全叠层的 20 秒受限版本,随后在 2026 年 4 月停止独立 Web 和 App 体验。这个序列说明,一个头部进入者可以多快地重新排序分发优先级, 或把产品折入更广的平台。在中国,相反方向也出现同一课。Moonshot 融资约 $2 billion,估值 $20 billion,据报道 ARR 超过 $200 million, 尽管它不是纯视频应用。资本正在流向通用实验室和智能体生态,它们可以把视频吸收为众多功能之一。如果买方越来越想要一个能写代码、搜索、推理并生成媒体的 AI 工作空间,独立视频领导者的增购会更难。 AIsphere 仍有可信路径,但比原始基准测试表现暗示的更窄。PixVerse 若要保持竞争力,需要同时守住三件事:第一,独立质量留在第一梯队; 第二,官方定价继续比高端西方同业更便宜或更简单;第三,公司通过模板、API、团队或地域合规匹配,赢下黏性的工作流位置。会削弱叙事的,是任何显示用户只是在近似相等的模型之间购物、 或把 Runway 和 Pika 这类聚合器当作控制平面的证据。在那种世界里,PixVerse 仍会作为强模型和应用存在,但利润率会流向别处。因此, 尽调负担是商业性的,不只是技术性的:投资人要承销持久竞争护城河之前,AIsphere 必须证明留存、输赢原因和企业合同黏性。[CP006, CP018, CP019, CP035, CP036, CP037]

护城河耐久性 / 竞争风险登记表
护城河主张或依赖竞争威胁严重性证据缓释措施 / 尽调问题
PixVerse 可以靠原始模型质量取胜Seedance、Kling、Vidu、Wan 和 Veo 的质量在基准场内高度聚集Artificial Analysis 排行榜显示多个相近替代品,且领跑者不是 PixVerse索取付费用户队列的标准化胜率、留存和输出偏好数据,而不只是基准截图
PixVerse 可以定价高于同业Runway、Seedance、Vidu 和开放权重 LTX 形成可见价格锚PixVerse 与 Runway 官方价格,加上 AA $/min 比较对照前五个替代方案,梳理实际 ASP、折扣和企业价格底线
独立应用分发仍然足够聚合器和路由层可以掌握订阅关系,同时切换底层模型Runway 打包第三方模型,Pika 承诺接入「所有模型」衡量客户多平台并用的频率,以及 PixVerse 是否能长期保持默认生成入口
中国与企业合规将成为差异化因素规则增加负担,但文档更完整的对手可以把合规转成销售优势中高中国标识规则、EU GPAI 义务和 NIST 信任预期都抬高了流程要求审查可见的治理材料、标识流程、版权立场,以及可直接给客户的信任文档
中国关联供应与扩展风险可控BIS 出口管制尽调仍可能让中国关联实体获取先进计算变复杂BIS 首页和 GT 指南确认许可要求仍在持续在出口执法收紧下,复核算力交易方、访问地域和备用容量计划
通用实验室短期内不是威胁Moonshot、Veo、Grok 和未来助手生态可以把视频吸进更宽的 AI 工作空间中高Moonshot 的融资规模,以及来自 Google/xAI 的基准测试竞品跟踪企业买家是否越来越偏好一体化 AI 套件,而不是单一用途视频应用

严重性衡量的是 AIsphere 持久竞争位置面临的风险,而不是 AI 视频需求整体是否有吸引力。核心问题是,价值留在 PixVerse,还是迁移到路由、操作系统或通用平台层。

[CP018, CP019, CP028, CP031, CP035, CP036]

3.5 图表

Chapter 04

04财务情况

4.1 定价与收入界面:有标价,但没有真实成交经济模型

不同于许多私有 AI 创业公司,AIsphere 确实暴露了可见商业界面。PixVerse 运营公开平台站、官方定价文档、消费者 App 流程,以及 2026 年产品更新, 后者描述团队计费、池化点数、API 访问和合作伙伴计划。最强的官方定价事实简单且有用:文档称在 Starter pack 下,$1 等于五段 V6 视频, 分辨率 720p、时长 5 秒、无音频。同一页面按模型、分辨率、时长、音频和运动模式计量使用,也展示更大的点数包和更高层套餐。这足以说明 AIsphere 已经从纯炒作进入真实点数经济。它还不足以回答投资人最终关心的问题。公开材料仍未披露真实企业定价、折扣、退款、免费转付费、 以及消费者订阅、预付点数、API 使用和谈判合同之间的分部收入结构。因此,抓取记录证明了变现意图和部分标价,但没有证明收入质量。[CI001, CI002, CI003, CI004, CI005, CI006]

收入流表
收入流机制单位当前价值 / 状态质量尽调问题
消费者积分或订阅支出用户通过 PixVerse 套餐和积分余额付费,在消费者端生成短片按月 / 按积分 / 按短片官方定价文档披露基于积分的标价和月度套餐结构,但实际付费用户转化未披露存在性可信度中,实际收益低提供付费用户数、转化率、退款率,以及按套餐拆分的活跃付费用户净收入。
预付积分充值客户购买一次性积分包,以增加生成量按包 / 按积分官方文档显示 $50 至 $5,000 的包档位,意味着钱包式追加销售打包方式可信度中,需求质量低披露充值频率、积分包相对订阅的收入占比,以及未用余额处理政策。
API 或开发者平台使用开发者或产品团队通过公开平台和 API 入口接入生成能力按次生成 / 用量合同平台页面和文档展示 API 定位,但实际 API 价格和用量未公开入口存在可信度中,经济性低提供 API 资费表、承诺用量条款,以及按模型家族拆分的毛利率。
团队工作区计费组织通过 Team Plan 购买共享积分、管理员控制和共享工作区按席位 / 共享积分 / 组织订阅官方生产平台更新确认合并计费和共享积分;未抓取到团队公开价格表功能存在可信度中,定价清晰度低提供 Team Plan 席位价格、超额规则和组织平均支出。
选择性 R1 伙伴合同工作室、流媒体平台和工具开发者通过受控伙伴准入接入实时视频议价合同 / 用量承诺官方伙伴计划文章称早期伙伴享受优惠定价且可沿用旧价,但没有披露公开数字GTM 路径可信度中,变现可见度低提供伙伴数量、平均合同价值、承诺用量和支持负担。
垂直工作流工具和商业制作服务Mini Apps 和工作流工具面向广告、营销和其他制作场景按活动 / 按工作流 / 企业合同官方工作流文章展示产品化商业工具,但未公开拆分分部收入低至中披露垂直工具是通过独立 SKU、用量提升,还是企业包扩张来变现。

各行区分公开可见的变现入口和仍未披露的实现数据。产品入口存在,并不证明它贡献了收入。

[CI001, CI004, CI005, CI006, CI007, CI008]
定价 / 变现表
入口价格 / 单位标价 vs 实际价格折扣 / 未知项来源
入门包 V6 标价锚点$1 = 5 条视频(V6,720p,5s,无音频)明确的官方标价锚点未知企业买家是否拿到显著不同的有效价格官方定价文档
分辨率、时长、音频和运动计费积分随模型参数变化;快速运动会让积分消耗翻倍明确的标价规则客户选择高阶设置的频率未知,套餐是否削弱定价也未知官方定价文档
预付积分包$50、$500、$2,000 和 $5,000 档位出现在抓取页面上可见标价促销、企业返利或年度承诺之后每美元可得多少积分未知官方定价文档
Scale 计划抓取文本显示价格为 $1,500,含 239,230 月度积分和 5,316 条生成视频可见标价,但页面抽取不完美抓取文本中入门层级版式部分错乱,因此对外引用前应实时复核低层级套餐官方定价文档
Business 计划抓取文本显示价格为 $6,000,含 1,069,500 月度积分和 23,766 条生成视频可见标价,但页面抽取不完美合同期限、用户上限、支持权益,以及显示的经济性是否对应特定模型基线均未知官方定价文档
Team 与伙伴合同基于报价或选择性定价;官方文章提到合并计费和伙伴旧价沿用,但未公开费率可能主要是实际定价,而不是公开标价未抓取到公开席位下限、折扣带、支持加价或承诺用量条款生产平台与 R1 伙伴计划更新

只复现清晰可读的官方数字。定价文档抽取有部分噪声,因此本表使用可见锚点,并在存在模糊处明确标注。

[CI002, CI003, CI004, CI005, CI006, CI007]
FI001: 收入模式桥

AIsphere 公开可见的收入路径,似乎从消费者创作和开发者接入,延伸到积分、团队、API 使用和议价合作合同;但实际收入和毛利仍不透明。

定性变现桥基于官方产品与工作流入口、以及 2026 年反复出现的收入和融资报道搭建,不是公司披露的分部收入瀑布。

[CI001, CI004, CI005, CI006, CI008, CI011]

4.2 GTM 代理指标与收入质量:漏斗很宽,但合同质量不透明

公开市场进入图景比单一消费者应用更宽。Google Play 文案、应用下载流程和 A16Z 消费者榜单仍指向创作者驱动的顶部漏斗; 而生产平台、Canvas 和 R1 合作伙伴计划更新显示,AIsphere 正在尝试把注意力转化为团队、开发者、工作室、流媒体平台和工具构建者。 Team Plan 的集中计费和池化点数等功能,是典型的组织增购机制;合作伙伴计划还明确具有筛选性,要求工程能力并提供路线图访问, 更接近企业销售,而非自助结账。增长代理指标有意义,但质量仍不均衡。2026 年 3 月报道反复呼应超过 $40 million ARR、超过 100 million 用户、超过 16 million MAU;公关稿和官方帖子又补充超过 175 个国家、生成视频 2 billion 条以上。这些数字让公司看起来商业上真实, 但没有回答多少客户实际付费、续约频率如何、ACV 分布怎样,或使用是否足够持久以支撑软件式留存。[CI008, CI009, CI010, CI011, CI013, CI014]

单位经济性表
指标数值 / 空值可信度为何重要尽调问题
官方入门包价格锚点$0.20 / 条 V6 720p 5s 无音频视频(由 $1 = 5 条视频推导)给出最清晰的公开单位价格底线,但没有反映实际组合或企业价格表提供按消费者套餐、企业合同和 API 工作负载拆分的实际 ASP。
公开 ARR>$40M(报告称)显示业务很可能已在有意义的规模上变现,但未说明收入有多持久、利润如何提供月度收入桥、递延收入和经审计师审阅的收入确认政策。
公开用户基数>100M 用户(报告称 / 公司声称)用户漏斗大,有助于解释投资人兴趣和免费增值式变现潜力按地域把用户拆成注册、活跃、付费和留存用户群。
公开月活跃用户>16M MAU(报告称)MAU 是有用的参与度代理指标,但不能替代付费账户或收入留存提供 MAU 到付费用户转化、付费用户流失,以及付费套餐 MAU 占比。
官方平台规模说法100M+ 用户、175+ 个国家、2B+ 条生成视频(公司声称)高使用量有助于数据飞轮和分发,但也可能抬高算力和审核成本披露付费与免费生成占比,以及每活跃用户的支持或审核成本。
创始人口径盈利标记订阅收入覆盖成本如果属实,这可能很重要,但该说法未经审计,也没有说明是否排除 R&D 或股权激励提供按主要产品线拆分的 GAAP 与 non-GAAP 贡献利润率。
毛利率 / NRR / 企业客户数这些是判断 AIsphere 是否像持久软件业务那样运转的核心质量指标提供毛利率、NRR、头部客户数、大客户门槛和同期群留存。

本表把公开牵引力代理指标和公共记录止步处的有意空值放在一起。空值标记缺失的承销输入,不是零值。

[CI002, CI011, CI012, CI014, CI015, CI021]
FI002: 单元经济性桥

公开证据在漏斗顶部的采用和可见积分定价上最强,在合同经济性上较弱,在留存与利润率质量上最弱。

桥的边界有意停在公开披露处。下游节点是尽调问题,不是已验证的公司指标。

[CI011, CI015, CI018, CI021, CI022, CI036]

4.3 成本结构与资本强度:更重的工作负载很可能把业务推上成本曲线

AIsphere 自己的材料暗示,利润率路径仍是核心财务问题。点数定价会随分辨率、时长、音频和更快生成而上升,这意味着面向客户的费率已经反映可变资源强度。 产品路线图随后进一步走向昂贵负载:面向合作伙伴的实时 720p API 访问、连续共享世界、多用户会话,以及实时 1080p 输出的官方说法。 这些都不能证明单位经济模型弱,但说明最高价值用例可能也是最吃算力的。上市公司文件在这里适合作类比,不可替代公司数据。Datadog 披露收入成本中的云托管成本, Cloudflare 披露网络和设备服务成本,Snowflake 警示最低云承诺,NVIDIA 则把每 token 成本描述为一场正在进行的基础设施战。 在这个背景下,AIsphere 没有披露毛利率、算力承诺或营运资本状况,是重要缺口。经济模型可能随更好的硬件和路由而改善,但公开证据尚未证明它们已经改善。[CI003, CI007, CI015, CI021, CI022, CI023]

FI004: 资本强度 / 现金流图

成本栈可能随更丰富的产品负载一路抬升,传导到云、网络、支持和承诺压力;最新一轮融资只是缓冲,而不是解决资金续航问题。

本图把公司产品信号与上市公司申报背景结合起来,映射可能的现金流压力点;它不是 AIsphere 披露的预算或烧钱瀑布。

[CI021, CI022, CI023, CI024, CI025, CI026]

4.4 资本充足性与融资依赖:新资金降低短期风险,但没有跑通现金跑道

AIsphere 并不像缺资本的预产品创业公司。多篇 2026 年 3 月报道独立把最新轮次定位为 $300 million,称其为中国 AI 视频赛道创纪录融资, 并称资金用于研发、新业务探索和全球扩张。AI Insider 又加入熟悉的独角兽框架,KrASIA 则称创始人认为订阅收入已覆盖成本。这些信号重要, 因为它们降低了即时生存风险,也显示投资人愿意为世界模型雄心提供资金。但承销上限仍低。抓取来源没有披露手头现金、月度烧钱、现金跑道、债务、 云承诺或下一轮触发条件。即便最强的正面证据也仍是方向性的,而非资产负债表级别。Runway 2026 年 2 月以 $5.3 billion 估值融资 $315 million,是有用背景:前沿 AI 视频同业即使证明需求后,仍可能需要非常大额融资。因此,AIsphere 最新融资是一层缓冲, 不是公司已经摆脱融资依赖的证明。[CI010, CI011, CI012, CI016, CI020, CI032]

资本充足性表
项目公开数值 / 状态可信度影响尽调问题
最新披露融资$300M Series C,2026 年 3 月为 R&D 和扩张提供有意义的短期资金缓冲获取交割备忘录、资金到账日期和投后股权影响。
资金用途表述R&D、新业务探索和全球扩张显示融资用于增长,而不只是修补资产负债表索取按算力、人员、销售和国际基础设施拆分的 12 至 24 个月预算。
投资人广度报告称 20+ 家机构;CDH 领投,境内外投资人参投广泛参与降低单一投资人风险,也可能支持未来融资通道提供董事会权利、按比例跟投义务,以及任何战略商业附函。
创始人口径经营覆盖KrASIA 称订阅收入覆盖成本方向信号有帮助,但不足以推断全公司盈利能力或现金生成提供经审计利润表,以及创始人表述中「成本」的定义。
手头现金没有现金数据,投资人无法估算偿付能力或战略灵活性提供最新非受限现金、受限现金和短期投资。
月度烧钱与资金续航这是判断融资依赖的最大缺失输入提供当前烧钱、情景资金续航,以及按工作负载拆分的月度算力支出。
债务、租赁和云承诺承诺会让看似资产轻的 AI 公司比表面上更接近固定成本结构披露债务、租赁、最低云承诺和硬件融资义务。
下一轮融资触发点投资人需要知道增长、模型训练或基础设施扩张是否很快会迫使公司再融资提供内部再融资触发指标,包括用户、收入和算力门槛。

公开证据只能证明融资渠道,不证明资本充足。缺失的现金、烧钱和承诺行,才是真正的尽调堵点。

[CI010, CI012, CI016, CI020, CI032, CI037]
FI003: 财务估计区间

最强的公开数字锚点是融资和报道中的 ARR;更大的背景则来自同业融资和上市公司积压订单指标,而不是 AIsphere 自身资产负债表披露。

所有数值均以百万美元计。同业项目用于说明资本强度和披露标准,不暗示 AIsphere 的估值或收入。

[CI010, CI011, CI027, CI032, CI037]

4.5 财务结论与尽调阻塞点:业务在变现,但还不能完全承销

财务结论复杂但清楚。正面:AIsphere 有可见标价、宽消费者漏斗、组织变现功能、反复出现的 ARR 和规模主张,以及非常新的 $300 million 融资,这应能买来执行时间。负面:证据仍停在决定收入质量和利润率耐久性的变量之前。公开材料没有展示客户数、 ACV 分布、队列留存、订单积压、递延收入、合同收入、毛利率、经营现金流、现金余额、债务,或按地域和客户类型的集中度。 上市可比公司让这种缺失更明显,因为它们通常披露这些标记。这不意味着 AIsphere 弱;它意味着投资人还无法区分一家强软件业务, 和一家快速增长但算力沉重、折扣沉重或服务辅助较多的业务。本章的正确姿态因此应保持谨慎:收入很可能真实,定价肯定真实, 但在管理层拿出董事会级财务资料和合同级队列证据前,收入质量仍是中低置信度。[CI027, CI028, CI029, CI030, CI031, CI033]

公开财务缺口表
缺失的私有指标对判断的影响确切尽调路径
按消费者、团队、API 和伙伴合同拆分的收入组合使人无法清楚判断集中度、质量,以及对低价值消费者使用的依赖索取按产品线和地域拆分的月度收入桥,并附付费账户数。
实际企业定价和折扣带使人无法把可见标价转成真实合同经济性复核前 20 大合同、资费表、折扣审批和超额收费表。
毛利率和算力 COGS使人无法判断实时 AI 视频是在盈利扩张,还是只是在拉高使用量索取模型级服务成本、GPU-hours、带宽、存储、审核和支持成本分摊。
客户集中度、ACV 和留存使人无法知道 ARR 来自持久软件账户,还是来自少数波动账户索取头部客户集中度、ACV 分层、同期群流失、NRR 和续约排期。
递延收入、积压订单或 RPO使人无法将 AIsphere 与公开软件式收入质量指标对比提供递延收入滚动表、已签积压订单和剩余履约义务。
现金、烧钱、资金续航、债务和云承诺使人无法对下一轮融资时点或下行韧性得出严肃结论索取当前资产负债表、债务明细、租赁义务和最低云承诺合同。

这些是截至 2026-07-03 阻碍对 AIsphere 财务形成清晰承销判断的最低公共数据缺口。

[CI027, CI028, CI029, CI030, CI034, CI036]

4.6 图表

Chapter 05

05产品与技术

5.1 已交付产品:PixVerse 现在覆盖消费者创作、协作工作流、可编程 API 和早期世界模型层

公开产品定义远不止一个提示词输入框。面向消费者的界面可见于 app.pixverse.ai,其导航如今暴露 Creation、Agent、Canvas、Mini-Apps、 Marketing Hub 和 API Platform;移动端教程和 Google Play 列表显示,同一套栈通过官方 iOS 和 Android App 分发,并具备云同步, 而不是一个独立缩水的伴侣应用。这很重要,因为它把 PixVerse 变成了跨界面的工作流:个人创作者可以在移动端从模板或文本提示词开始, 团队可以转入共享工作区和集中资产,开发者可以把同一生成工作流升级为 API 或 CLI 自动化。2026 年生产平台更新把这一层次显性化: 增加 Team Plan 权限、池化计费和 Mini Apps;Canvas 则为脚本、参考、分镜、批量和成片提供节点式看板。按工作流看,AIsphere 交付的是 AI 视频的创作者操作系统,而不只是模型端点。战略含义偏正面:工作流所有权和界面宽度,比原始排行榜比较更能提高切换难度。但需要注意的是, 公开证明最强的是打包和功能广度;每一层已经承载多少收入或企业留存,证据要薄得多。[CE001, CE002, CE003, CE004, CE005, CE006]

产品模块 / 资产矩阵
模块 / 入口主要用户状态 / 成熟度差异化尽调缺口
消费者创作 App + Web个人创作者和半专业创作者成熟公开入口文本、图像、转场、延展、参考和模板跨模式创作速度快公开材料未按 Web 与 App 用户群拆分付费转化或留存
Canvas 视觉工作区创作者、代理商和内部团队新近推出,但已明确产品化节点式画板把脚本、参考、批量产出和最终成片留在一个可追踪工作区未找到按团队规模拆分的公开使用或采用指标
Team Plan + 共享资产库协作制作团队已商业发布共享工作区、基于角色的权限、集中资产和共享计费,把 PixVerse 推出单用户工具范畴席位价格、管理员控制和审计日志深度未公开
Mini Apps / Ad Master营销人员和垂直场景运营者早期产品延伸面向特定场景打造的工作流外壳,缩短从创意简报到交付物的距离只有第一个 Mini App 有详细描述;更广模块路线图仍然稀疏
PixVerse Platform API + CLI 路径开发者、工作室和企业活跃但部分受控可编程生成,配有标价积分和自动化钩子公开文档仍以定价为主;完整 API schema、SLA 和认证细节不可见
V6 / C1 自研视频模型高端视频创作者和制作团队核心商业引擎自研模型栈支持 1080p 输出、音频、镜头控制和按秒定价独立证据支持一线质量,但不能证明无人挑战的品类领导地位
R1 实时世界模型工作室、平台开发者、互动体验开发者新兴 / 选择性准入持续生成可互动世界,把 PixVerse 推到短片合成之外全面开放时间、吞吐量和可靠性仍未披露
IP / 伙伴工作流层权利方、游戏工作室、品牌项目实验性但有辨识度KAGAMI Gate 授权和 Tripo look-dev 创造了并非所有对手都开放的工作流钩子现有证据停留在伙伴博客深度,而不是长期制作案例研究

状态反映截至 2026-07-03 公共入口可见内容:成熟意味着已广泛上线,已发布意味着商业入口存在,新兴意味着伙伴受控或证据仍少。

[CE003, CE005, CE006, CE007, CE009, CE012]
工作流 / 使用场景表
用户任务当前工作流PixVerse 方案可衡量收益限制
用文本提示词或图像创建短社交视频单用途移动编辑器或云端短片生成器消费者 App/Web,支持文生视频、图生视频、模板和延展一个账户贯通 Web 和移动端,快速输出 HD积分经济性和重试次数仍会影响可用性
运行多镜头品牌或广告工作流分散的提示词工具、文件夹、电子表格和编辑器Canvas 加 Mini Apps 和 Team Plan 工作流入口项目状态、资产和变体留在一个画板上,而不是散在多个标签页创意审批和企业管理员控制的公开证据有限
在产品或流水线内自动化生成手动 Web 生成或拼接第三方脚本API Platform 加上声称的 CLI 支持可编程生成,并有透明的按秒 / 积分定价公开文档尚未披露完整 SLA、认证或配额
原型化互动世界或实时叙事游戏引擎加定制资产和逻辑栈R1 实时世界模型和伙伴 API连续世界生成、虚拟形象、同步音频和叙事操控选择性伙伴准入意味着该入口并非广泛自助
在 3D 建模前验证游戏资产外观静态概念图之后盲建模PixVerse 电影感 look-dev 连接 Tripo Studio团队可以在网格制作前测试光照、运动和轮廓证据是一篇伙伴工作流文章,不是中立客户案例研究
推出授权粉丝或品牌模板非官方粉丝内容,或事后下架风险KAGAMI Gate 受控授权加 Captain Tsubasa 模板权利方条款在创作时适用,而不是发布后才处理授权系统仍是概念验证,且只针对特定 IP
运营移动优先的创作者工作流浏览器优先工具,手机支持弱官方 iOS/Android App 加云同步移动中生成和直接社交分享扩大分发App 仍由云端支撑,而不是本地运行,因此服务质量仍取决于后端性能

只有在所审阅来源显示具体工作流变化、定价逻辑或产品控制差异时,才列出收益;这些不是经审计的 ROI 说法。

[CE004, CE007, CE008, CE009, CE018, CE019]
FE002: 客户工作流 / 运营流程

典型 PixVerse 工作流如今从想法捕捉进入有组织的项目状态,再到生成、审阅,最终外部发布或集成。

[CE004, CE007, CE008, CE009, CE010, CE018]
FE004: 产品成熟度 / 能力图

消费者创作和工作流包装看起来更成熟;R1 和公开信任披露仍呈现较早期特征。

成熟度标签是分析师判断,综合公开产品广度、接入模式和治理披露得出,不是内部推出指标。

[CE003, CE005, CE007, CE009, CE018, CE019]

5.2 架构与运行模式:底层是自研视频基础模型,R1 则推向持久多模态世界引擎

公开架构故事有两层。第一层是围绕 V6、V5.6、C1 和相关创作模式的成熟视频生成栈。官方发布材料、定价文档和详细 V6 评测显示, 这是一个云端模型家族,已经支持文本生成视频、图像生成视频、转场、延展、参考生成视频、集成音频,以及 1080p 输出,并配有透明点数定价。 第二层是较新的 R1 方向,PixVerse 不把它描述为短片生成器,而是实时世界模型:连续交互式 1080p 视频、omni-native 多模态处理、 低采样步数、记忆增强注意力,并更新支持单张照片 avatar、持久会话和多用户共享世界。这一运行模式在战略上重要,因为它把 AIsphere 从只比拼短视频质量, 推向可能支撑模拟、交互娱乐和实时环境的架构。但合作伙伴计划的打包显示,这一层还不成熟:最具新意的 R1 能力仍是选择性开放的早期访问基础设施, 不是广泛自助的生产 API。换句话说,自研视频基础模型层今天看起来已经商业化;世界模型层技术上很有野心,但仍部分处于普遍可用前阶段。[CE010, CE011, CE012, CE013, CE014, CE015]

技术 / 运营架构表
层 / 组件角色依赖风险
消费者创作外壳承载创作模式、模板、参考素材和创作者体验Web 应用、移动应用、账号同步和云端生成后端功能面很宽,仍可能掩盖企业级可观测性或治理披露不足
工作流编排层存储项目状态、节点、参考素材、故事板、批次和团队资产Canvas、Team Plan、共享素材库和权限模型PixVerse 从短片工具走向协作型记录系统,运营复杂度随之上升
自研视频基础模型生成核心 V6/C1 短片输出,覆盖镜头、音频和多镜头行为训练数据、推理基础设施和按秒计费引擎独立基准显示质量强但并非领先无争议,原始模型优势仍可被挑战
音频 / 场景控制层协调语音、音乐、音效、镜头运动和镜头连续性集成音频开关、字面化提示词规范和多镜头处理品牌准确的产品镜头、多语言对白和混乱动作仍需要人工复核
R1 世界模型层生成连续可交互的视频世界,而不是一段段离散成片Omni 原生多模态处理、低步数采样、记忆增强注意力选择性开放的包装意味着真实吞吐量和支持能力披露仍不足
API / 自动化层让合作伙伴和开发者用脚本生成,或嵌入流水线平台定价文档、声称的 CLI 访问和 R1 合作伙伴计划公开文档更强调商业访问,对实现细节或 SLA 深度披露较少
权利 / 合规层在分发或创作时执行标识、IP 规则和使用政策中国标识义务、KAGAMI Gate 控制、应用商店审核、内部审核官方页面未描述类似信任中心的机制或溯源标准
算力 / 供应链层供给训练和推理所需的 GPU 与先进计算出口管制环境、供应商关系和后端云容量BIS 指引显示,与中国相关的算力获取仍是现实运营依赖

这张架构表把软件模块、治理层和外部依赖放在一起,因为 PixVerse 的公开运营模式是商业工作流加云服务,不只是狭义机器学习里的模型架构。

[CE009, CE010, CE012, CE014, CE015, CE016]
FE001: 产品架构图

PixVerse 把创作者 UX、工作流编排、自研生成模型和初现轮廓的实时世界模型层,叠在云交付和合规控制之上。

[CE003, CE005, CE009, CE012, CE015, CE018]

5.3 部署、集成与差异化:护城河更多来自工作流所有权、生态钩子和选择性伙伴分发,而不是绝对基准测试领先

顺着部署界面往外看,AIsphere 的差异化最容易理解。消费者通过模板、参考图工具、语音和动作控制,以及跨平台 App 分发进入。 团队随后可以进入 Canvas 看板、共享资产库和基于角色的协作。开发者可以通过 API 平台自动化,并且据公司自述材料还可使用 CLI 访问。 最后,更复杂的合作伙伴可以通过面向工作室、流媒体平台和工具构建者的门控计划使用 R1。围绕 Tripo Studio、KAGAMI Gate、Captain Tsubasa 和 AI for Good 的生态帖子显示,AIsphere 正在尝试把 PixVerse 嵌入游戏资产视觉开发、权利管理 IP 工作流、品牌模板生态和全球创作者项目。 这是一个真实产品策略,但并不意味着 PixVerse 是无争议的技术领导者。Artificial Analysis 仍把 V6 放在第一梯队,而不是绝对榜首; AppBrain 评论显示,部分用户仍遇到漂移、不想要的编辑或点数浪费。结果是一个平衡的产品结论:PixVerse 对 AI 视频创业公司来说异常宽、 商业打包也成熟,但公开证据更有把握支撑“具备强工作流设计的生产平台”,而不是“明确的类别最佳模型”。[CE020, CE021, CE022, CE024, CE025, CE026]

路线图 / 发布 / 开发阶段表
日期 / 阶段功能 / 里程碑状态含义来源
2026-01R1 作为实时 AI 世界模型发布已发布 / 早期访问确立世界模型方向和交互式 1080p 定位官方发布帖
2026-02PixVerse V5.6 围绕独立排名做营销已发布并推广显示产品对盲测基准定位有信心官方排名帖 + Artificial Analysis
2026-03V6 发布,带来镜头控制、角色表演、多语言文字、多镜头音频和 CLI 集成已发布把 PixVerse 从快速短片工具推进到面向生产的工作流官方 V6 发布
2026-03生产平台扩展,纳入 Team Plan、Mini Apps、CLI 和 API 叙事已发布买家从单人创作者扩展到团队和开发者生产平台帖子
2026 年春发布 Canvas 可视化工作区已发布围绕生成增加项目状态管理和批量编排Canvas 帖子
2026-04AI for Good 峰会工作坊和电影节项目生态里程碑强化负责任、全球创作者平台定位AI for Good 帖子
2026-06开放 R1 720p 实时 API 和合作伙伴计划选择性开放最新世界模型能力有了商业相关性,但仍需通过合作伙伴门槛R1 合作伙伴帖子
2026-06R1 更新加入单张照片头像、持久会话和多人世界现版本升级R1 更接近协作环境,而不是演示会话R1 更新帖
2026-06 至 2026-07KAGAMI Gate / Captain Tsubasa 上线与挑战赛概念验证部署显示 PixVerse 不只能做通用粉丝内容,也能交付授权且受治理的 IP 工作流KAGAMI + Captain Tsubasa 帖子

路线图表同时纳入产品发布和生态部署,因为 PixVerse 的产品论点既依赖核心模型迭代,也依赖围绕这些模型包装起来的工作流和 IP 项目。

[CE005, CE007, CE011, CE015, CE017, CE018]
FE003: 关键依赖图

PixVerse 依赖云算力、监管合规、合作伙伴渠道和应用商店 / 社区分发,不亚于依赖模型质量本身。

[CE021, CE031, CE033, CE034, CE036, CE037]

5.4 信任、合规与局限:监管可见,但公开治理披露仍比产品界面薄

公开信任证据好坏参半。正面看,PixVerse 明显运行在可见治理约束内:中国 2023 年生成式 AI 规则适用于公开视频服务, 2025 年标识措施要求对 AI 生成内容做可见和基于元数据的标记;KAGAMI Gate 实验也说明,公司正在测试带许可的 IP 授权层, 而不是只靠下架。Google Play 和 AppBrain 界面也让移动端权限足迹可读,Harris 和 BIS 材料则强调合规风险不是假设: 内容标识执法、出口许可和工作流文档都会影响运行。弱点在于公司没有公开展示的内容。覆盖官方 Web、App、博客和平台界面后, 仍看不到公开信任中心、SOC 2 或 ISO 27001 页面、公开可用性 / 状态历史,也看不到某些前沿同业已发布的水印和来源证明披露。 这种缺失不能证明内部控制弱,但意味着审核、滥用处理、水印、API SLA、事件响应和 GPU 供给韧性,仍必须从公开网页核查转入私下尽调。 对承销来说,这是主要技术风险:产品广度已经公开,企业级治理和可靠性还没有。[CE030, CE031, CE032, CE033, CE034, CE035]

信任 / 质量 / 合规表
控制 / 信号状态范围缺口
中国生成式 AI 服务规则公开适用面向中国公众提供的视频服务必须使用合法数据、保护个人信息、标识特定输出,并提供稳定服务PixVerse 的合规落地细节没有公开文档
2025 年中国 AI 内容标识规则公开适用且有时间约束对 AI 生成的文本、图像、音频、视频和虚拟场景,要求显式标识以及元数据 / 隐式标识PixVerse 公开页面没有说明用户或 API 实际获得哪种标识路径
KAGAMI Gate 的 IP 权利管控可见概念验证为 Captain Tsubasa 合作提供受权利管理的模板创作框架有潜力,但尚未在多个大型 IP 系列中验证
移动端权限与隐私足迹可从应用商店和 AppBrain 页面看到相机、麦克风、存储、通知、广告 ID 和网络访问,与媒体生成应用相符用户仍需要更清楚的公开说明,涵盖数据处理、广告和留存实践
公开可靠性 / 信任中心页面已审阅官方页面未发现状态历史、SOC 2、ISO 27001、事故报告、信任中心或 API SLA公开控制平面很薄,企业买家仍需私下尽调
公开溯源 / 水印披露已审阅 PixVerse 材料未发现生成视频是否带有明确溯源元数据或默认水印前沿同行披露更多,PixVerse 的信任差异化仍不清楚
用户可见的输出质量和点数投诉参差AppBrain 评论称赞语音模型和免费点数,也提到漂移和点数浪费有投诉不等于系统性失败,但说明生产可靠性并非毫无摩擦

“未发现”指已审阅公开页面没有出现该项;不证明内部不存在该控制。

[CE031, CE032, CE033, CE034, CE035, CE036]

5.5 图表

Chapter 06

06客户情况

6.1 买方、用户和付款方分层横跨消费者创作者、团队和嵌入式用户

PixVerse 的公开客户地图在用户上很宽,在已披露付款方上窄得多。最可见的用户是消费者创作者,他们通过移动 App、Google Play、 第三方 Android 商店、Product Hunt 和病毒式模板页面接触 PixVerse。这些界面针对快速社交视频创作优化,而不是采购式销售。 第二层在其上方:专业消费者、营销人员和机构,他们需要可重复的活动工作流、客户组织和更好的输出质量。官方 Team Plan 和 Canvas 材料显示, PixVerse 明确在为这一层产品化,提供共享工作区、池化点数、基于角色的权限,以及活动或客户组织。再往外,公司还通过 R1 合作伙伴计划面向 API 和嵌入式用户;那里的买方是工作室、平台开发者或工具构建者,而不是消费者创作者。净结果是可信的买方—用户—付款方阶梯, 但公开证据仍更清楚地标出用户界面,而不是谁在有意义规模上付款。[CU001, CU002, CU005, CU007, CU008, CU009]

客户分层表
分层买家 / 用户 / 付费方主要用例公开证明页面战略价值关键缺口
消费者创作者个人创作者通过自助套餐发现、使用,有时付费快速社交短片、照片动画、文生视频、病毒式特效Google Play、应用下载页、AppBrain、Uptodown、APKPure最大的可见漏斗顶端用户基数付费转化和订阅用户数未披露
半专业创作者 / 营销人员个人或小团队付费;创作者或营销人员是日常用户营销活动短片、产品演示、适合社交平台的品牌内容APKPure 面向营销人员的话术、Canvas 工作流文案、热门模板页面把随手创作接到重复商业使用分层收入、ACV 或留存拆分没有公开
团队 / 代理商工作区管理员付费;创作者、审核者和管理员使用产品共享资产管理、营销活动组织、合并点数、审批生产平台的 Team Plan 和 Canvas 页面最清晰的新兴企业式层级没有具名代理商、席位数或续约证明
API / 嵌入式合作伙伴工作室、平台开发者或工具构建者付费;下游终端用户消费输出嵌入产品或工作流的实时视频生成R1 合作伙伴计划和平台页面自助订阅之外潜在 ACV 更高的路径选择性项目,未披露合同规模或在用客户数
IP / 授权合作伙伴品牌或 IP 合作方加终端粉丝;商业付费方组合不清楚授权模板库和受治理的角色 IP 使用Captain Tsubasa 合作和 KAGAMI Gate PoC让 PixVerse 区别于通用提示词工具推广价值可见,经常性收入条款不公开
机构 / 创作者项目取决于项目设计,组织者、赞助方或创作者都可能付费电影节投稿、工作坊、社区叙事UN AI for Good 项目和 Product Hunt 社区页面把信任和触达延伸到纯娱乐应用之外这些项目的变现和续约未披露

分层来自公开产品页面和具名项目推断;证据对用户的识别比对高价值付费方更清楚。

[CU001, CU005, CU007, CU008, CU010, CU029]
FU001: 客户旅程图

公开证据显示,路径从自助发现走向付费接入、团队工作流、合作伙伴集成,随后进入仍未解决的续约阶段。

[CU001, CU008, CU021, CU023, CU029, CU031]

6.2 采用证据最强的地方是应用商店分发和广泛创作者触达

AIsphere 的公开采用证据,在消费者平台发布遥测或排名的地方最强。AppBrain 给出最清晰的外部读数:页面显示 50,000,000+ 下载量, 同时又称该 App 已被下载 73 million 次,过去 30 天大约 980 thousand 次下载,在摄影类排名 #2,并基于约 4.3 million 个评分给出 4.49 分。Product Hunt 补充了一个小得多但仍有用的专业消费者信号:304 名关注者、6 条评论和 4 次发布。Apptopia 确认了 MotivAI Private Limited 名下的 iOS 上架条目;Google Play、Uptodown 和 APKPure 合在一起显示,分发并不局限于单一应用商店。 公司主张层面,PRNewswire、PixVerse 生产平台更新和 2026 年 3 月融资报道都强化了同一标题:超过 100 million 用户、国际触达和非常大的累计视频量。 这些数字没有揭示的是,用户中有多少会付费、留存,或扩展成更高价值合同。[CU012, CU013, CU014, CU015, CU016, CU017]

客户增长 / 采用轨迹表
指标日期来源置信度含义缺失分母
AppBrain 安装基数快照页面显示 50,000,000+ 次下载;生命周期下载 73M2026-07 快照AppBrain外部可见的 Android 分发足迹很大商店 / 下载数不等于付费用户
AppBrain 近期势头过去 30 天下载 ~980K 次;摄影类 #22026-07 快照AppBrain2026 年中消费者漏斗看起来仍活跃没有同期群、地域拆分或付费转化
AppBrain 评分基数约 4.3M 条评分给出 4.49 / 52026-07 快照AppBrain大规模消费者互动足迹,也可粗略代理总体正向情绪评分不等于留存、收入或企业满意度
Product Hunt 半专业用户足迹304 名关注者;6 条评论;4 次发布;4.3 分2025-12 归档快照Product Hunt(经 Wayback)低至中显示移动商店之外存在一定创客 / 半专业用户认知相比大众应用安装量,社区足迹偏小
公司新闻稿规模口径100M+ 用户和 800M+ 条生成视频2025-09-24PR Newswire2026 年 3 月融资周期前已存在大规模安装 / 用户基数未披露付费占比或活跃用户频率
融资报道规模口径100M+ 用户;16M MAU;$40M+ ARR2026-03CnTechPost / Yicai / AI Insider多篇 2026 年 3 月报道复述同一套大规模叙事用户和 ARR 口径未按分层或地域拆分
官方平台规模更新覆盖 175+ 个国家的 100M+ 用户;已生成 2B+ 条视频2026-07 快照PixVerse 生产平台更新公司同时声称覆盖广度和产出量加速没有关联到付费账户数、续约或账户集中度
iOS 分发存在应用由 MOTIVAI PRIVATE LIMITED 上架,分类为照片与视频 / 娱乐2026-07 快照Apptopia低至中显示 PixVerse 是多平台,不只 Android捕获页面未公开 iOS 下载量或收入估计

轨迹表混合市场遥测、公司口径和融资报道,因为 AIsphere 不发布常规付费客户 KPI 看板。

[CU012, CU013, CU014, CU015, CU016, CU017]
FU002: 采用 / 部署流程

公开路径从广泛发现和激活,走向更小规模的付费、协作和合作伙伴主导用例。

[CU012, CU017, CU021, CU022, CU025, CU027]

6.3 具名验证确实存在,但更靠合作伙伴和模板牵引,而不是典型企业大客户背书

最具体的公开具名验证,并不是带量化 ROI 的 Fortune 500 客户故事。它们更多是合作伙伴计划、品牌活动、工作流集成和创作者项目。Captain Tsubasa 和 KAGAMI Gate 说明,PixVerse 已经成为授权 IP 活动和 IP 使用追踪概念验证的操作界面。Tripo Studio 则给出一个具体的游戏美术流程:在构建 3D 资产之前,PixVerse 先承担电影感视觉开发环节。UN AI for Good 合作让 PixVerse 获得带工作坊和电影节曝光的机构级创作者项目。与此同时,模板页也重要,因为它们揭示普通用户的真实行为:用户使用 Dancing Baby、Winter Sovereign、Fly to the Sun 等预制、低门槛创意格式,而不只是从空白画布做提示词工程。这是有意义的客户证据,但更适合解读为活跃社区行为和合作伙伴试验,而不是多元化企业收入的证明。[CU021, CU022, CU023, CU024, CU025, CU026]

具名客户证明表
客户 / 合作伙伴分层部署或用例生产还是试点结果 / 证据质量局限
Captain Tsubasa / KAGAMI GateIP 授权 / 粉丝创作授权角色模板和受治理的 AI 视频 IP 使用追踪已上线限时活动加概念验证基础设施具名交易对手、具体用例,并出现在官方应用商店文案和官方活动页面未披露合同经济性、续约或收入贡献
UN AI for Good Global Summit 峰会机构创作者项目联合国舞台上的工作坊和 AI for Good Film Festival 投稿2026 年已上线项目具名机构合作伙伴和全球可见的创作者行动号召项目声誉清晰,直接变现不清楚
Tripo Studio游戏 / 3D 工作流合作伙伴在转为 3D 资产前,先用 PixVerse 做视觉开发已上线集成功能游戏团队的具体待办任务明确,比泛泛联合营销更强没有具名下游付费工作室或合同细节
R1 合作伙伴计划API / 嵌入式用户面向工作室、平台开发者和工具构建者选择性早期开放早期商业项目买家类型和商业机制明确客户未具名,定价仍是选择性开放,而非标准 GA

这些是具名公开证明,但大多是合作伙伴 / 项目案例,不是已披露的经常性企业客户。

[CU027, CU028, CU029, CU030, CU031, CU032]
社区模板与用例行为表
模板 / 行为受众公开证据面复用或变现信号局限
AI Dancing Baby以社交为先的消费者创作者PixVerse 热门模板页公司称获得数百万次社交观看,并支持一键复用创作官方页面未披露独立用户数或花费
Winter Sovereign奇幻 / cosplay / 影视感创作者PixVerse 热门模板页说明转换工作流可复用,不止于单一梗图格式没有使用次数或活动转化数据
Fly to the Sun 模板影视化叙事创作者PixVerse 热门模板页说明模板驱动的图生视频叙事可在不同用户间复用未按模板家族披露留存或收入
Captain Tsubasa 授权特效动漫 / 足球粉丝和品牌活动用户官方活动页加 Google Play 列表页授权模板被推到消费者获客入口,而不是只藏在企业材料里限时活动可能拉出尖峰,而不是沉淀长期用户群
邀请码和 Discord 行为早期采用者和活跃社区成员官方 R1 邀请码指南持续社交分发和 Discord 活动说明,高级功能仍在反复吸引关注社区参与不等于付费留存或企业价值

这些行展示围绕模板和社区运营的公开行为模式;它们可作为采用度代理指标,但不是收入质量的直接证明。

[CU022, CU025, CU026, CU027, CU032]
FU003: 客户证据矩阵

有具名交易方和具体工作流的地方,证据质量最高;但所有公开细分场景的收入和持续性可见度都偏弱。

单元格概括的是公开证据深度,不是内部销售状态;它们区分证据哪里具体、哪里仍不透明。

[CU015, CU027, CU030, CU031, CU034, CU038]

6.4 重复使用的代理信号存在,但留存和耐久性披露仍然薄弱

PixVerse 确实有用户回访的公开信号:付费订阅者可自动获得 R1 访问权限,高分辨率被标为高质量付费输出,邀请码持续通过 Discord 和社交渠道发放,模板页也在多个主题上刷新,并未停留在上线时的静态物料。消费者满意度代理指标同样方向为正,AppBrain 的评分基数很大,Product Hunt 的评论集较小但也提供支持。但这些代理指标没有一项达到客户耐久性的尽调门槛。它们没有告诉投资者有多少用户转为付费、留存多久、总留存或净留存是否为正,品牌项目是否转化为重复支出。即使是最强的官方变现页面,也只停留在定价界面、访问规则和协作功能。因此,本章把评分、模板刷新和订阅门槛视为重复使用线索,而不是留存队列或续约数据的替代品。[CU014, CU021, CU022, CU023, CU024, CU033]

留存 / 重复使用 / 满意度表
指标值 / 状态分层置信度尽调要求
付费客户数所有已变现分层提供付费订阅用户数、付费团队数和付费 API 合作伙伴数。
NRR / GRR / logo 流失团队、API 和企业相邻账户提供按分层拆分的同期群留存桥,列出收缩、流失和扩张。
合同期限 / 续约条款Team Plan、API 合作伙伴、品牌项目提供合同期限中位数、续约窗口和续约率。
客户集中度 / 头部账户占比企业和合作伙伴收入提供 top-1、top-5、top-10 客户收入占比,以及合作伙伴集中度。
公开满意度代理AppBrain 约 4.3M 条评分给出 4.49 / 5;Product Hunt 6 条评论给出 4.3消费者创作者 / 半专业用户用付费同期群 CSAT 或 NPS,把活跃用户满意度与安装基数情绪分开。
重复使用代理付费 R1 访问、邀请码社区和持续更新的模板库,暗示存在持续使用消费者创作者 / 半专业用户按同期群披露订阅用户留存、付费月活和重复生成频率。

null 表示在已审阅来源集中未公开披露;非 null 行只是代理指标,不是真正的留存或收入质量指标。

[CU014, CU015, CU021, CU022, CU023, CU024]

6.5 团队、API 和合作伙伴扩张路径都说得通,但集中度和收入结构仍未知

从公开材料看,PixVerse 的扩张逻辑清楚。广泛的创作者分发带来自助试验;付费 R1 访问和更高质量设置形成向上销售路径;Team Plan 和 Canvas 补上机构、营销团队和专业创作者需要的控制能力;合作伙伴计划则为工作室、平台开发者和工具构建者提供嵌入式或 API 路径。Captain Tsubasa 和 AI for Good 等品牌项目也说明,PixVerse 可以把产品打包进授权和营销活动场景。这是一条可信的先落地再扩张故事。未解决的问题是集中度。纳入审阅的来源都没有披露企业 / API 收入是否已经具备实质规模、是否集中在少数合作伙伴手里,或消费者订阅是否仍主导收入结构。KrASIA 的融资报道也再次提醒:即使管理层称订阅收入覆盖成本,公开客户经济性仍然单薄。投资者看得到动作,但仍无法为账户基础做承销判断。[CU039, CU040, CU041, CU042, CU043, CU044]

扩张与集中风险表
扩张驱动公开证明集中风险影响尽调路径
免费创作者向付费订阅用户转化付费订阅用户可立即使用 R1,官方指南强调高级质量免费到付费转化未披露如果转化健康,可能成为消费者变现的核心杠杆按套餐披露转化率、付费方 ARPU 和订阅用户留存。
单人创作者转向团队工作区Team Plan、共享工作区、合并点数和 Canvas 项目组织未公开具名团队、席位数或续约把消费者端动能接到更高价值账户的关键桥梁披露团队版客户 logo、席位增长和续约同期群。
团队工作流到 API / 嵌入式项目面向工作室、平台开发者和工具开发商的选择性 R1 合作伙伴计划选择性准入可能意味着收入集中在少数伙伴身上可能是 ACV 最高的路径,也是最大的不透明来源披露伙伴数量、合同规模和头部伙伴敞口。
品牌 / IP 活动Captain Tsubasa、KAGAMI Gate 和 AI for Good 表明,PixVerse 能打包品牌化或使命驱动项目可能只是一次性推广收入,而非可重复收入在授权和活动业务上增加战略可选项披露复购项目数量、合同结构和续约历史。
工具链 / 游戏工作流伙伴Tripo Studio 拼出一条从准专业创作者到专业用户的具体工作流尚未公开具名的下游付费工作室支撑游戏和 3D 扩张,但商业深度尚未验证披露伙伴带来的活跃账户和工作室扩张情况。
分发平台和社区渠道Google Play、iOS、第三方商店、Product Hunt 和社交邀请码渠道带来发现流量获客组合可能过度依赖平台和病毒式传播漏斗很大,但平台政策和潮流风险更高按商店、网页、推荐和伙伴渠道披露获客结构。

风险行基于公开渠道结构和披露缺口;它们描述的是可能成立的扩张机制,不是已经验证的实际收入结构。

[CU021, CU039, CU040, CU041, CU042, CU043]

6.6 图表

Chapter 07

07风险

7.1 按严重程度排序的风险概览

AIsphere 的风险栈由三个可能相互叠加、而非孤立出现的问题牵引。第一,PixVerse 是一个同时面向中国和国际界面的公开 AI 视频产品,因此中国的生成式 AI 和标识规则不是理论背景:它们直接约束内容标识、投诉处理、个人信息义务,以及监管认为控制不足时可能触发的服务暂停。第二,前沿视频质量和低延迟推理仍依赖先进计算和云基础设施,使 AIsphere 暴露在美国出口管制收紧、NVIDIA 供应限制,以及 Alibaba Cloud 支撑的全球部署集中度之下。第三,公司处在整个生成式 AI 行业尚未解决的版权地带,训练数据来源、输出所有权和授权 IP 使用仍有争议。下一层风险是模型质量和消费者分发脆弱性,因为 AI 视频里的应用商店规模和榜单地位可能快速逆转;再往下是资本和执行风险,因为 AIsphere 刚拿到充足融资,但在烧钱速度、客户集中度和合规人员配置上几乎没有公开披露。因此,风险图景更适合被理解为一条传导链:监管和供给先传导到产品可靠性、企业信任、利润率质量,最终传导到估值。[CR036, CR041, CR042, CR043, CR045]

FR001: 风险热力图

AIsphere 主要风险集群的可能性、影响、缓释成熟度和剩余暴露。

单元格综合有来源支撑的严重性和缓释成熟度;它们是判断性排序,不是概率预测。

[CR041, CR042, CR043, CR045]

7.2 法律与监管风险

法律与监管暴露是 AIsphere 最清晰的第一层风险,因为 PixVerse 正好集合了监管想要约束的特征:面向公众的生成式视频、实时交互、用户上传内容,以及不断扩大的 API 分发。中国暂行办法适用于向公众提供、生成文本、图像、音频、视频或其他内容的生成式 AI 服务,要求训练数据来源合法,并通过深度合成制度明确规定标识义务。2025 年标识办法进一步加入显式面向用户的标记和隐式元数据要求,同时推动提供者完善投诉处理、用户协议、日志记录和内容安全控制。对 AIsphere 来说,这不是遥远政策标题,而是具体落地负担,因为 PixVerse 同时经营中国和全球网页界面、消费应用和生产工作流。出口管制是第二条主要法律线索:BIS 在 2026 年澄清,先进计算物项仍需为中国相关实体申请许可证,因此任何前沿 GPU 访问收紧都可能直接传导到模型节奏和服务经济性。第三条法律线索是版权。美国生成式 AI 诉讼仍然没有解决训练数据合理使用、输出所有权和下游责任问题;AI 视频产品在营销品牌或角色类输出时,还会面对额外暴露。AIsphere 已开始尝试授权 IP 基础设施,这是方向上的加分项,但公开证据尚未显示出成熟、经过审计的法律控制栈。[CR001, CR002, CR003, CR004, CR005, CR006]

监管 / 法律风险登记表
风险司法辖区 / 制度状态发生概率严重程度缓释成熟度剩余敞口尽调路径
中国生成式 AI 与标识合规CAC 暂行办法 + 2025 年标识规则已生效极高索取 CAC 备案或安全评估历史、标识实施日志和投诉处理流程
前沿 GPU 出口管制准入BIS 先进计算管制许可要求已生效极高审查 GPU 供应商、许可状态、库存覆盖和备用硬件计划
训练数据和输出版权敞口美国及全球 AI 版权诉讼未解决审查训练数据来源、赔偿条款和输出权利政策
欧盟透明度和文档义务EU AI Act / GPAI 下游使用分阶段落地按地域梳理 API、客户和文档义务
个人信息和用户记录处理中国数据和用户保护义务已生效检查留存、删除、DPA 和用户请求响应控制
授权 IP 治理品牌内容和授权基础设施概念验证审计权范围、续约条款和侵权升级流程

严重程度按监管直接抓手,以及风险向服务连续性、企业信任和估值传导的力度排序;本表具有重大性,但不是穷尽式法律审查。

[CR001, CR003, CR004, CR007, CR009, CR014]

7.3 运营、模型质量与安全风险

从运营上看,AIsphere 已不再只是一个消费者特效应用。公开材料现在把它描述为生产平台、实时世界模型功能和 API 合作伙伴计划,这扩大了延迟、审核、合作伙伴支持和企业可靠性上的故障面。实时视频在这里尤其关键,因为连续交互比一次性短片生成更难标识、审查和审核,尤其当用户提示词、上传内容和下游分发都进入规模化之后。公开证据支持的是强劲动量,而不是成熟运营披露:Artificial Analysis 显示 PixVerse 位居领先视频模型厂商之列,AppBrain 显示其移动端采用规模很大,公司也强调功能覆盖广。缺失的信息同样重要。公开来源没有显示扎实的事件记录、企业 SLA 表,或中国及国际市场合规运营的实施证据。对云和基础设施供应商的依赖也值得关注。Cloudflare、Datadog 和 Snowflake 等可比公有云公司的公开申报文件都警告,宕机、安全事件和第三方基础设施故障会损害可用性和客户信任。AIsphere 不太可能免于这些动态;如果说有区别,快速产品扩张和实时视频反而把运营门槛抬得更高。结果就是:可靠性、审核和安全今天并非已知弱点,但相对于产品路线图的野心,公开记录明显不足。[CR011, CR012, CR013, CR016, CR018, CR019]

运营 / 质量 / 安全风险登记表
失效模式发生概率严重程度缓释成熟度剩余敞口未解决缺口
GPU 或云容量中断极高供应商条款、容量缓冲和备用计划未披露
实时标识或审核失效没有公开证据证明实时视频控制基准或审计结果
服务宕机或延迟恶化中高没有公开的企业 SLA 表或事故历史
安全或隐私事件中低没有公开审计包、泄露历史或详细数据治理材料
模型质量相对头部玩家倒退中高没有独立披露胜率或失败率的节奏
应用商店分发或支付中断商店排名依赖和抽成经济性未披露

本表结合公司披露,以及规模化云与软件平台在公开文件中反复提示的风险模式。

[CR013, CR016, CR031, CR032, CR033, CR034]
FR002: 风险传导图

监管、基础设施和质量三类失灵如何传导到收入、利润率和估值。

边展示下行情境可能传导的方向,而不是完全量化的因果模型。

[CR033, CR046, CR048]

7.4 合作伙伴、依赖、客户与竞争风险

AIsphere 的依赖栈异常密集。据报道,Alibaba Cloud 为训练、推理、全球部署和合规提供全栈支持,这让双方关系具有战略价值,也让云集中度成为直接的经济和可用性风险。GPU 依赖进一步放大问题,因为前沿视频模型仍需要先进计算,而世界模型野心推高算力需求时,出口规则正好可能影响中国相关访问。分发也很集中。应用商店规模是优势,但也意味着发现、支付、排名可见度和一部分客户信任都落在第三方平台上,而不是 AIsphere 自有渠道里。企业变现引入了另一种依赖模式:R1 API 合作伙伴计划是选择性开放,而非完全开放,因此近期企业收入可能依赖一组相对较小的合格合作伙伴和支持资源。授权 IP 与工作流合作同样有两面性。Captain Tsubasa 和 KAGAMI Gate 显示 AIsphere 正在构建权利清晰的增长循环,Tripo 集成拓宽了创作者工作流,但每个合作伙伴也都可能成为延误、重新谈判或稀释的节点。竞争进一步抬高风险,因为 Runway 和其他视频实验室都在讲类似的世界模型叙事;如果质量领先地位滑落,品类领导者的快速轮换可能把今天的牵引力变成明天的流失。[CR020, CR021, CR022, CR023, CR024, CR025]

伙伴 / 依赖风险登记表
依赖项交易对手角色集中度失效场景严重程度缓释剩余敞口
先进 GPU 供应NVIDIA / 受出口管制的供应链核心模型训练和推理容量损失、价格飙升,或前沿算力获取延迟极高硬件多元化和许可纪律
云部署骨干Alibaba Cloud 及相关基础设施供应商训练、推理、部署、合规支持价格冲击、宕机或战略优先级调整会伤及可用性和利润率多区域架构和第二供应商规划
移动应用分发Google Play / 应用商店生态获客、计费、信任界面中高商店政策、排名或下架会打击下载和变现网页分发和更强的直连渠道中高
企业 API 变现选择性 R1 伙伴分发和早期企业转化入驻慢或伙伴流失会拖慢企业收入结构转移更广的自助入驻和伙伴赋能
授权 IP 增长回路KAGAMI Gate 和权利方品牌模板和权利管理权利失效或执行薄弱会造成活动中断或法律敞口合同约定的权利范围和可审计内容跟踪
工作流生态扩张Tripo Studio 和创作者工具伙伴面向创作者和工作室的延展用例中低伙伴路线图延误会削弱差异化工作流价值建设原生能力或冗余集成

共同模式是:外部伙伴带来增长、基础设施或权利覆盖,同时也变成战略脆弱点。

[CR024, CR025, CR026, CR027, CR028, CR029]
FR003: 依赖关系图

围绕算力、云、分发、权利和生态扩张的关键外部依赖。

该图强调依赖集中度,而不是合同层级或收入分成。

[CR025, CR027, CR035, CR047]

7.5 资本与执行风险

即便完成 Series C,资本风险仍然真实存在,因为新融资没有回答核心承销问题。公开证据显示 AIsphere 融资 $300 million 并推进更大的产品界面,但烧钱速度、现金跑道、客户回本、头部客户集中度和留存后毛利率结构仍未披露。这很重要,因为 PixVerse 的定价明确以积分为基础、门槛低;它可以加速采用,也可能掩盖一个问题:在完整分摊算力和支持成本之后,重度使用能否转化为耐久的贡献利润率。执行风险也高于公司增长叙事暗示的水平。组织现在需要一支管理梯队,同时支撑合规、信任与安全、合作伙伴赋能、企业支持、云运营和法律监督。但公开材料大多强调产品速度,而不是这些职能背后的人员深度。结果是一个中高风险项:融资实力降低了短期破产担忧,但运营和财务披露不完整,投资者仍无法检验 AIsphere 能否把消费者级规模动量转化为耐久、合规且经济性有吸引力的企业增长。[CR017, CR030, CR038, CR039, CR043]

人员 / 执行风险登记表
角色 / 职能依赖或缺口发生概率严重程度缓释尽调路径
财务和规划烧钱速度、跑道和单位经济模型披露缺失月度规划纪律和保守流动性目标审查董事会材料包、月度现金预测和情景模型
法务 / 信任与安全运营中国和全球合规工作量在扩张,但看不到匹配的人手细节专门的政策、审核和法务运营梯队索取组织架构、按职能划分的员工数和升级处理 KPI
企业支持和解决方案工程选择性 API 计划意味着入驻能力稀缺伙伴手册、SLA 责任归属和支持团队配置检查伙伴入驻漏斗和响应时间数据
管理梯队深度公开材料更强调产品速度,而不是管理层冗余中高补入有经验的财务、合规和企业 GTM 运营人才审查高管履历、董事会观察员和招聘计划
模型评估和红队测试实时视频扩大了失效模式,需要有纪律的测试正式发布前评估、对抗测试和事故演练索取评估量表、发布闸门和复盘流程

执行风险的核心不在创始人质量,而在支持、合规和财务职能是否跟得上产品野心的扩张速度。

[CR017, CR030, CR038, CR039, CR043]

7.6 缓释措施、监控指标、推翻投资论点的触发因素与尽调问题

可见的风险缓释故事连贯但不完整。AIsphere 并没有忽视风险:授权 IP 试验、选择性 API 合作伙伴计划、企业工作流工具和公开基准测试,都说明公司在尝试从病毒式消费者新奇产品,转向具备权利意识、可供企业使用的基础设施。NIST 式治理原则为公司最终应展示的控制能力提供了可信蓝图。但当前证据基础仍指向一种混合而非已证明的缓释成熟度。因此,投资者应监控少数硬信号,而不是宽泛叙事:PixVerse 能否在实践中证明中国标识和投诉处理合规,前沿 GPU 或云访问是否稳定,应用商店评分和下载动量是否守得住,企业合作伙伴转化是否走出选择性试点,以及未来披露是否终于界定烧钱速度和客户集中度。最清晰的推翻投资论点事件也很明确。中国监管行动、失去可信的先进算力访问,或质量和分发持续滑坡、导致消费者牵引力无法转化为企业收入,都会实质削弱投资案例。最高优先级尽调问题是合规材料、GPU 和云合同、客户集中度数据、烧钱速度和现金跑道,以及法律、信任与安全、合作伙伴支持的人手深度。[CR044, CR045, CR046, CR047, CR048]

缓释与终止标准表
风险可监控触发项阈值 / 事件行动含义
中国合规和标识监管通知或审计证据缺失下架、警告,或无法展示标识工作流暂停投资测算,直到控制措施得到验证
前沿 GPU 准入许可延迟或容量缓冲收窄所需算力的可信可用期少于 6-12 个月下调增长和模型迭代节奏假设
版权和来源没有训练数据控制,或赔偿条款薄弱管理层无法证明来源、权利已清理数据或赔偿安排降低企业收入置信度,或回避
云集中度单一供应商依赖仍占主导面对主供应商,没有实质性故障切换或商业议价能力折减可靠性和毛利率假设
应用商店依赖消费者分发走弱评分、下载量或商店状态显著恶化下调规模和变现假设
模型质量预期相对头部玩家的基准表现滑坡多个发布周期失去一线质量地位下调护城河和转化假设
资本充足性跑道缩短快于计划没有已承诺资本且跑道不足 12 个月把融资视为投资逻辑的关键条件
执行深度支持或合规梯队没有扩张伙伴积压、未解决事故,或合规招聘缺口若未纠正,转为继续研究

这些触发项设计成可监控、与投资相关,而不是抽象风险标签。

[CR044, CR045, CR048]

7.7 图表

Chapter 08

08估值

8.1 融资背景和隐含 ARR 倍数

AIsphere 有足够证据支撑估值讨论,但还不足以支撑一个干净价格。多篇 2026 年 3 月报道相互印证了 $300 million C 轮,并重复同一组商业摘要:AIsphere 的 ARR 已超过 $40 million、用户超过 100 million、月活用户超过 16 million。对一家 AI 视频创业公司来说,这些都是有意义的规模标记。它们也符合早前报道的框架:公司在本轮之后方向上已经是独角兽级资产。问题在于,纳入保留的公开来源没有完全对齐准确的投后估值、股权结构影响,或所售证券的经济性。 这种不确定性很重要,因为它会改变名义倍数的解读。如果公司估值约为 $1 billion,且 ARR 只是超过 $40 million,隐含名义倍数约为 25x ARR 或更低,因为披露的 ARR 数字只是下限。表面看,这比最热门的 AI 视频和中国 AI 稀缺性可比公司温和。但折价并不显然就是便宜。它主要补偿的是收入质量、付费客户结构、毛利率和优先权结构上的低可见度。换句话说,AIsphere 真实到足以留在漏斗里,但透明度还不足以把当前名义价格视为自证合理。[CV001, CV002, CV003, CV004, CV005, CV006]

建议摘要表
决策字段当前观点支持证据决策含义
建议继续研究公司看起来真实且已有规模,但公开证据仍未充分说明收入质量和安全条款。让 AIsphere 继续留在漏斗中;当前价格暂不放行投资。
置信度融资、产品和可比公司证据都真实存在,但具体估值机制和 ARR 质量仍缺少足够文件支持。优先做数据室尽调,而不是直接否决或给出正面投资条款建议。
风险评级估值支撑取决于收入转化、算力经济性,以及 AI 资本市场窗口是否有利。任何交易都应按高尽调强度、对下行情景敏感来处理。
估值立场偏高约 $1B 的公开估值若对应 >$40M ARR,倍数低于更热同行,但也反映披露和验证更弱。要么拿到更扎实证据,要么争取更宽松入场价格。
入场纪律若股权结构表和收入桥接不清,不按当前估值投入新资金最大未解变量是优先股堆叠、付费客户质量和毛利率耐久性。争取价格保护、里程碑分期,或者等待。
上调路径经审计证据或更低入场价更清晰的 ARR 桥接和条款包,比又一条叙事型融资新闻更能快速改变判断。只有证据或价格出现实质变化时再重估。

本表刻意对价格敏感:它总结的是当前估值已经隐含的假设,而不是抽象地给 AIsphere 打分。

[CV007, CV026, CV030, CV031, CV032, CV033]
FV001: 推荐逻辑

真实规模信号被较弱的收入质量和条款单可见度抵消,因此得出这一推荐。

该流程是定性的,不是概率化的;它映射保留估值证据所暗含的决策链。

[CV001, CV003, CV007, CV026, CV030, CV033]

8.2 可比公司组,以及 AIsphere 实际所处位置

解读 AIsphere 隐含估值的最好方式,是把它放在一个情绪亢奋但内部分层的可比公司组中。Runway 是最接近的私有 AI 视频工作流可比公司:TechCrunch 称其在 2026 年 2 月以 $5.3 billion 估值融资,而 Sacra 估算其 2025 年年化收入为 $90 million,隐含约 59x 收入倍数。Pika 低得多,估值约 $470 million,但公开来源没有足够清晰地披露收入,无法得出可靠倍数。高端则是 Moonshot 的 2026 年 5 月融资和 4 月 ARR 披露,隐含大约 100x ARR 下限;MiniMax 的香港 IPO 估值相对其披露收入和亏损,也说明只要叙事动量还在,公开市场愿意接受非常高的中国 AI 倍数。 按表面算术,AIsphere 比这些更热的可比公司便宜,但它的验证也更弱。独立基准测试把 PixVerse V6 放在有竞争力的第一梯队,而不是明确主导位置;品类仍足够拥挤,买方切换是真风险。公开记录也缺少后期软件或基础设施投资者通常会要求的企业级披露。因此,正确结论不是「AIsphere 很便宜,因为 25x 低于 59x 或 100x」。更好的结论是,市场可能有充分理由给 AIsphere 一个披露折价,同时承认公司确实有真实规模和可信产品。[CV008, CV009, CV010, CV011, CV012, CV013]

可比估值表
可比对象关键指标倍数 / 估值 / 状态相关性局限
AIsphere(私有,2026 年 3 月)>$40M ARR 口径;>100M 用户;>16M MAU方向上为独角兽级 / 约 $1B 公开估值;准确投后估值未披露本次直接评估标的ARR 是公司关联口径的下限,证券条款未公开。
Runway(私有,2026 年 2 月)Sacra 估计 2025 年年化收入为 $90M估值 $5.3B;按 Sacra 估计约 59x 收入最接近的 AI 视频工作流和世界模型可比公司,且 2026 年融资可见收入来自第三方估计而非审计,且 Runway 的企业端验证更广。
Pika(私有,2024 年 6 月)消费端优先的 AI 视频产品;收入未披露$80M 融资后估值约 $470M可作为该品类底部区间的低端 AI 视频可比对象估值时间较早,也缺少干净收入基数来做倍数比较。
Moonshot AI(私有,2026 年 5 月)2026 年 4 月 ARR 超过 $200M约 $20B 估值;按披露下限粗略约 100x ARR有最新 ARR 口径的中国 AI 稀缺性上限可比对象模型、API 和智能体范围远比 AIsphere 更宽。
Zhipu AI(香港 IPO,2026 年 1 月)香港 IPO 目标募资约 $640M约 $6.7B / HK$51.2B 的 IPO 估值目标说明 2026 年公开市场投资人愿意支持中国 AI 稀缺资产我们保留的样本提供的标准化收入细节有限,难以干净比较倍数。
MiniMax(香港 IPO,2026 年 1 月)9M25 收入 $53M;亏损 $512MIPO 估值约 $6.5B,随后在模型驱动热情下市值涨至 $33B最适合观察泡沫化后期风险偏好的中国公开 AI 视频相邻可比对象上市日和上涨后的定价波动很大,不能代表稳态公允价值。

这是一组样本型可比对象,目的是在私有 AI 视频同行和中国公开 AI 发行人之间框定估值逻辑,而不是强行得出一个“正确”倍数。

[CV002, CV007, CV008, CV011, CV013, CV015]
FV002: 估值敏感性

相比又一条叙事型融资头条,AIsphere 的价值对证据质量和条款更敏感。

数值是 1 到 5 的序数型影响评分,不是概率,也不是机械定价模型。

[CV020, CV022, CV023, CV024, CV025, CV037]

8.3 投资论点、反论点与建议

投资论点很直接。AIsphere 不是只停留在路演 PPT 的公司。它融到了真实资本,开放了真实产品界面,展示了真实用户规模主张,也出现在独立基准测试集合里,而不是只活在自家营销材料中。这些事实重要,因为它们把 AIsphere 与尚无收入的概念故事区分开来,也让独角兽级叙事至少方向上说得通。反论点同样重要。公司没有展示准确的估值机制,ARR 主张仍与公司相关而非审计确认,现有证据也没有告诉投资者收入基础主要是粘性的企业 / API 支出,还是更脆弱的创作者订阅与推广流量组合。Adobe、Snowflake、Datadog、Cloudflare 和 NVIDIA 的公开申报文件,让这个披露缺口无法被忽视。 在这些抵消因素下,可支撑的判断是继续研究,而不是买入。信心应为中等,因为公司质量案例真实存在,但证券质量案例仍不完整。风险应为高,因为估值支撑取决于多件事同时做对:收入转化、算力纪律,以及仍然打开的 AI 资本市场窗口。因此,估值立场应为偏紧。按名义估值看,AIsphere 并不显然荒谬,但当前证据基础太薄,不能说价格有吸引力。[CV019, CV020, CV021, CV022, CV023, CV024]

投资逻辑 / 反向逻辑表
论点方向支撑证据改变判断的证据
AIsphere 是已有规模的真实资产,不是概念故事。投资逻辑公司已融资 $300M,开放付费产品入口,并出现在独立基准测试样本中。经审计的分渠道收入和队列数据,会把这条从可信度证据推进到完整承销证据。
当前估值仍可能主要是叙事估值。反向逻辑公开记录尚未对准确投后估值、优先股堆叠或确认收入质量做出对账。干净的投资条款清单和收入桥接会降低叙事溢价。
从表面算术看,AIsphere 比最热 AI 可比公司便宜。投资逻辑粗略 25x ARR 上限低于 Runway 和 Moonshot 的公开倍数。如果 ARR 质量偏弱或条款激进,这个折价就会消失。
PixVerse 有竞争力,但尚未清楚占据主导。反向逻辑Artificial Analysis 显示,它是在竞争紧密的第一梯队,而不是遥遥领先的基准王者。如果能长期领跑基准测试,或拿出分场景胜率证据,溢价倍数论据会更强。
披露质量明显低于成熟上市软件公司的规范。反向逻辑上市公司文件会披露收入质量和利润率标记,而 AIsphere 没有公布。管理层级 KPI 披露和审计报表会缩小可信度折价。
证券条款质量可能落后于公司质量。反向逻辑看不到公开股权结构表、清算瀑布或附函经济条款。干净条款提升可投性的速度,可能快过更多增长标题。

反向逻辑主要针对定价、证据质量和证券结构,而不是否认 AIsphere 有真实产品。

[CV019, CV020, CV021, CV022, CV026, CV029]
FV004: 投资 KPI

AIsphere 在战略相关性和产品可信度上得分高,但在证据深度和证券可见度上得分低。

评分是基于运行日已保留证据的 IC 式启发判断,不是管理层 KPI。

[CV019, CV020, CV026, CV029, CV030, CV031]

8.4 乐观 / 基准 / 悲观情景逻辑与价格纪律

在这里,情景分析比单点目标更站得住。在乐观情景下,AIsphere 会把公开 ARR 和用户主张转化为更清晰的审计桥,证明企业 / API 经济性具备实质意义,并让 PixVerse 保持足够竞争力,使公司继续受益于板块层面的稀缺性。在这种事实组合下,约 $1.4 billion 到 $1.8 billion 的估值可以支撑。基准情景下,公开商业故事方向上真实,但还无法干净承销:ARR 真实但部分偏创作者,利润率仍难评估,条款仍是私有信息。这支撑大约 $0.85 billion 到 $1.1 billion 的估值,使当前名义估值更接近公平偏紧,而不是显然便宜。 悲观情景是双重打击。如果付费转化、企业质量或 API 深度令人失望,同时 AI 视频稀缺性溢价降温,估值可能压缩到大约 $0.55 billion 到 $0.75 billion。这并不要求公司失败。它只需要投资者决定,一家仍不透明的 AI 视频创业公司不应再按最亢奋同行的方式定价。因此,入场纪律比叙事热情更重要:升级路径是更好的证据或更好的价格,最好两者都有。[CV033, CV034, CV035, CV036, CV037, CV038]

乐观 / 基准 / 悲观情景表
情景假设估值 / 回报逻辑概率信号关键风险
乐观经审计 ARR 和渠道结构验证其具备软件质量的经常性收入;条款干净;PixVerse 在基准测试中保持第一梯队。$1.4B-$1.8B;只有披露质量追上故事时,当前估值才显得可接受。低-中需要企业 / API 经济性比当前公开记录显示得强得多。
基准ARR 被证明真实,但收入结构仍只披露一部分,证券条款继续不公开。$0.85B-$1.1B;今天的估值大致在合理到偏高之间,安全边际有限。投资人仍暴露在倍数压缩和条款清单意外之下。
悲观在 AIsphere 展示更清晰经济性之前,付费转化、企业收入质量或行业风险偏好转弱。$0.55B-$0.75B;公司即便经营未失败,也可能被重估到远低于独角兽叙事的水平。中-高叙事溢价和收入质量担忧可能同时破裂。

区间是情景估值,不是 DCF 输出;公开证据对股权结构机制、确认收入质量和利润率结构披露不足,不能假装精确。

[CV033, CV034, CV035, CV036, CV037, CV038]
FV003: 估值 / 回报区间

采用情景区间比单点目标更站得住脚,因为价格支撑取决于仍然缺失的证据。

区间是基于已披露 ARR 下限、可比公司估值区间和披露质量推导的编辑性情景估计,不是 DCF 或退出模型式精确测算。

[CV033, CV034, CV035, CV036, CV038, CV042]

8.5 最终尽调问题和推翻投资论点的触发因素

剩余尽调工作异常集中,而且高度可执行。第一,投资者需要按渠道拆出从订单额到 ARR 再到确认收入的月度桥,因为当前 >$40 million ARR 主张不足以说明 AIsphere 是否配得上软件质量倍数。第二,投资者需要 2026 年 3 月条款清单、投后股权结构表、清算瀑布和所有附函,因为在同一个名义估值下,证券质量可能远弱于公司质量。第三,投资者需要单位经济性证据:每视频分钟算力成本、按产品界面划分的毛利率,以及收入中类似续约而非类似营销活动的比例。 终止触发因素也同样具体。降价轮、沉重的高级清算优先权,或薄弱的收入桥都会说明独角兽叙事跑在了证据前面。基准测试差距扩大或付费转化降低,会削弱产品质量能够支撑溢价经济性的主张。香港或中国 AI 融资环境转冷,则会拿走目前支撑可比公司的稀缺性溢价。在这些事项关闭前,正确姿态是继续尽调,而不是确信买入。[CV037, CV038, CV040, CV041, CV042, CV043]

打破投资逻辑与终止触发项表
触发项阈值对投资逻辑的传导行动含义
降价融资或重结构化融资新资金定价低于隐含独角兽水平,或加入激进的优先级条款说明当前公开估值高估了外部支持,或高估了证券质量。除非条款大幅改善,否则从继续研究转为回避 / 等待重置。
收入桥接薄弱从质量看,确认收入、ARR 构成或 API / 企业端结构支撑不了 >$40M 的公开口径打破“规模信号正在转化为持久软件经济性”的投资逻辑。将估值重定价到悲观区间。
付费客户质量薄弱低付费转化、续约弱,或用户规模背后高度集中让产品热度变成比估值假设更弱的变现故事。要求更低入场价或里程碑式结构。
基准测试滑坡PixVerse 失去第一梯队位置,或在质量 / 成本权衡上明显落后削弱产品竞争力能够支撑溢价定价的论据。降低乐观情景概率,并压缩可接受倍数。
行业倍数压缩香港或私募市场对中国 AI 的风险偏好明显降温移除当前支撑私有和新上市可比公司的稀缺性支撑。从尽调转为等待重新定价。

这些是与入场价格和证券质量相连的估值专项终止标准,不是对一般风险登记表的重复。

[CV027, CV028, CV040, CV041, CV042]
最终尽调清单表
主题缺失证据重要性负责人 / 尽调路径
收入桥接按消费者、API 和企业渠道拆分的月度订单额、ARR、递延收入和确认收入判断 >$40M ARR 口径到底支撑软件质量倍数,还是只支撑促销式规模故事。财务团队、董事会材料和审计师资料包。
股权结构表与优先股堆叠C 轮后股权结构表、清算瀑布、反稀释条款和附函经济条款界定当前公开估值下的证券质量和真实下行风险。公司法律顾问和融资文件审查。
单位经济性按产品入口拆分的毛利率、每视频分钟算力成本、支持负担和云承诺检验 AI 视频使用量能否带着利润扩张,而不只是看起来有规模。FP&A、基础设施和云合同审查。
队列质量付费客户数、NRR、流失、扩张,以及头部客户集中度区分用户规模和可持续经常性收入质量。收入运营和账单导出。
企业 / API 证据具名企业客户、API 收入占比、合同条款和续约历史判断 AIsphere 是否应更接近工作流 / API 同行估值,而不是消费端新奇工具。销售负责人加头部客户合同抽样。
下一轮 / 退出路径下一轮融资、老股交易或上市计划的时间和结构当前私有市场估值部分受有利的 AI 融资和香港发行窗口支撑。董事会、领投方和银行顾问。

这些问题按其改变推荐结论或可接受入场价的速度排序,而不是按获取难易度排序。

[CV037, CV043, CV044, CV045, CV046]

8.6 图表

免责声明

本报告仅供参考,反映截至 2026-07-03 的公开来源尽调。AIsphere 是私人公司;许多商业、治理和证券细节仍未经审计或未披露,任何投资决定前都应独立核实。

证据索引

结论
编号陈述可信度来源
CO001 AIsphere is the English name used by Beijing Aishi Technology Co., Ltd., whose official corporate site describes it as an AI-video-generation model and applications company. SO001, SO020
CO002 Public registry-style and encyclopedia sources place AIsphere’s establishment on 2023-04-07 in Beijing. SO020, SO014, SO017
CO003 AIsphere’s headquarters are publicly described as being in Beijing. SO014, SO015, SO020
CO004 The company’s official positioning is to build world-leading AI video generation large models and applications for the AGI era. SO001
CO005 PixVerse is AIsphere’s flagship AI-video product surface for global users across web and mobile. SO002, SO003, SO024, SO025
CO006 Founder and CEO Wang Changhu previously worked at Microsoft Research Asia and led visual-technology efforts at ByteDance. SO021, SO014, SO017
CO007 Co-founder Xie Xuzhang is quoted publicly on financing strategy and product direction in 2026 coverage. SO010, SO011
CO008 AIsphere launched PixVerse for overseas users in January 2024 and released a domestic beta shortly afterward. SO014
CO009 The company completed more than CNY 400 million of Series A fundraising by March 2025, including an A5 round led by Eminence Ventures. SO015
CO010 Yicai reported that AIsphere employed around 50 workers as of March 2025. SO015
CO011 PixVerse had more than 40 million worldwide users and more than 15 million monthly active users by March 2025, according to Yicai. SO015
CO012 Alibaba led AIsphere’s $60 million Series B round in September 2025. SO016, SO017, SO009
CO013 Series B participants publicly named by TMTPOST included Fortune Capital, Shenzhen Capital Group, the Beijing AI Industry Investment Fund, Giant Network, and Antler. SO017
CO014 Public 2025 company statements said PixVerse had already passed 100 million users worldwide and more than 800 million generated videos. SO016
CO015 AIsphere’s March 2026 Series C raised $300 million and set a new record for a single China AI-video financing round. SO009, SO010, SO011
CO016 CDH Investments led the March 2026 Series C round, with more than 20 institutions participating according to CnTechPost. SO009
CO017 Yicai said Series C investors also included Ruyi Holdings, 37 Interactive Entertainment, E-Town Capital, Suzhou Capital Group, Lion X Fund, and UOB Venture Management. SO010
CO018 Late-2025 to early-2026 public coverage put AIsphere’s annual recurring revenue above $40 million. SO009, SO010
CO019 Yicai reported that PixVerse and Paiwo AI together exceeded 100 million users, with more than 16 million monthly active users, by October 2025 / early 2026. SO010
CO020 AIsphere publicly launched PixVerse R1 in January 2026 as a real-time world model supporting 1080p generation. SO006, SO007, SO018
CO021 Official product materials describe R1 as moving AI video from offline clip generation to continuous, interactive world simulation. SO006, SO007, SO002
CO022 AIsphere launched V6 in March 2026 with stronger camera control, character performance, multilingual text rendering, multi-shot audio, and CLI integration. SO005, SO004
CO023 Sina reported in December 2025 that AIsphere had iterated its self-developed video model five times in about two years. SO019
CO024 PixVerse’s official homepage says its enterprise stack combines proprietary video foundation models, APIs, template ecosystems, and near-real-time generation for production workflows. SO002
CO025 Official product pages market PixVerse across both consumer creation and enterprise API use cases. SO002, SO003, SO024
CO026 Official and company-linked sources frame V5.6 as ranking #2 globally on Artificial Analysis video-generation leaderboards. SO008, SO016
CO027 Sina reported that PixVerse also entered a16z’s global top-50 generative-AI consumer mobile-app ranking at #25 in 2025. SO019, SO022
CO028 The Google Play listing shows PixVerse distributing frequent consumer-facing features such as text-to-video, image-to-video, upscaling, templates, and licensed effects. SO023
CO029 PixVerse maintains dedicated app-download and mobile tutorial pages for iOS and Android distribution. SO024, SO025
CO030 AIsphere and Alibaba Cloud announced a full-stack AI cooperation in December 2025 covering infrastructure, model services, products, ecosystem, and commercial globalization. SO019
CO031 The Alibaba Cloud agreement explicitly tied cloud infrastructure and compliance support to training and inference for PixVerse’s proprietary video model. SO019
CO032 KrASIA reported that Wang Changhu told 36Kr subscription revenues from AIsphere’s products already cover costs. SO013
CO033 KrASIA also framed AIsphere’s 2023 founding as a contrarian bet made while many investors doubted independent video-model startups could survive against OpenAI and Chinese internet giants. SO013
CO034 Public reporting repeatedly places AIsphere in a highly competitive field that includes OpenAI Sora, ByteDance Seedance/Jimeng, and Kuaishou Kling. SO009, SO013, SO014
CO035 Official materials present AIsphere as a vertically integrated AI-video company spanning proprietary models, consumer creation tools, and enterprise APIs rather than a single demo model. SO001, SO002, SO003, SO024
CO036 The public record still does not disclose a full board roster or a clean, current cap-table breakdown. SO020, SO021, SO009
CO037 The exact March 2026 post-money valuation is directionally around unicorn level but not independently reconciled in the fetched public record. SO009, SO010, SO011, SO012
CO038 User, ARR, and “largest-in-sector” claims are concentrated in company-linked or single-chain media reports rather than audited disclosures. SO009, SO010, SO013, SO016
CO039 AIsphere appears to be following a consumer-first global distribution strategy before deeper enterprise monetization. SO013, SO019, SO024, SO025
CO040 The fetched public evidence supports AIsphere as a real Beijing-based AI-video startup with fast funding momentum, but leaves governance, exact valuation mechanics, and independently audited economics unresolved. SO001, SO009, SO010, SO020, SO021
CM001 TBRC defines generative AI in video creation as the use of generative models to create, edit, or enhance video content. SM002
CM002 TBRC segments the category by deployment, applications, and end users, specifically naming on-premise versus cloud, marketing/education/entertainment/social media, and large enterprises/SMEs/individual creators. SM002
CM003 Research and Markets separately frames the market by deployment, application, end user, and end-use industries such as media and entertainment companies and educational institutions. SM001
CM004 Official vendor surfaces show that the market boundary now spans consumer creation, production workflows, and APIs rather than only one-off text-to-video clips. SM008, SM013, SM015, SM016
CM005 Runway, Pika, Vidu, Kling, Jimeng, and Wan all market AI video generation directly, confirming a crowded substitute set outside AIsphere. SM008, SM024, SM025, SM026, SM027, SM028
CM006 Upstream GPU/training spend and the broader generative AI software stack are better treated as adjacent or enabling spend than direct revenue for video-creation vendors. SM001, SM002, SM003
CM007 TBRC says the generative AI in video creation market reached $0.39 billion in 2025, will grow to $0.47 billion in 2026, and is projected to reach $0.98 billion in 2030. SM002
CM008 Research and Markets says the generative AI in video creation market is valued at $0.47 billion in 2026 and projected to reach $0.98 billion by 2030 at a 20.4% CAGR. SM001
CM009 The two narrow-market publishers agree that dedicated generative-video-creation revenue remains sub-$1 billion today even while growing around 20% annually. SM001, SM002
CM010 Fortune Business Insights sizes the broader generative AI market at $103.58 billion in 2025, $161 billion in 2026, and $1,260.15 billion by 2034. SM003
CM011 Fortune says North America held 48.70% of the broader generative AI market in 2025 and that enterprise adoption plus foundation-model innovation are major growth drivers. SM003
CM012 Because the broader generative AI market is roughly two orders of magnitude larger than the narrow video-creation estimates, using one figure as a stand-in for the other would materially overstate direct addressable spend. SM001, SM002, SM003
CM013 a16z says AI video models moved from experimental to fairly dependable for short clips over the prior six months. SM004
CM014 The same a16z report says Hailuo, Kling, and Sora debuted on the web rankings and that Hailuo and Kling surpassed Sora in monthly visits by January 2025. SM004
CM015 Artificial Analysis frames category competition around quality Elo, speed, and pricing across text-to-video, image-to-video, and audio-enabled models. SM005
CM016 Artificial Analysis' text-to-video leaderboard lists PixVerse V6 at Elo 1,069 and $6.90 per minute, versus Veo 3.1 at Elo 1,096 and $24.00 per minute and Vidu Q3 Pro at Elo 1,082 and $9.60 per minute. SM006
CM017 Artificial Analysis' image-to-video leaderboard lists PixVerse V6 at Elo 1,074 and $6.90 per minute, versus Veo 3.1 at Elo 1,086 and $24.00 per minute. SM007
CM018 Benchmark dispersion is narrow enough on quality but wider on price that workflow fit and cost control matter almost as much as absolute leaderboard position. SM005, SM006, SM007
CM019 Google Play and PixVerse's blog/home surfaces position PixVerse for simple creator workflows, viral effects, and mobile-first experimentation. SM011, SM017
CM020 TBRC explicitly lists marketing and social media among the main applications for generative video AI. SM002
CM021 Vidu markets fast workflows for social content, ads, and storytelling, showing how SMB marketers can buy AI video for campaign production rather than entertainment alone. SM025
CM022 PixVerse's production-platform update says teams need collaboration, asset management, automation, and distribution, indicating that media or IP buyers care about more than clip generation quality. SM015
CM023 PixVerse's blog homepage highlights officially licensed Captain Tsubasa templates, suggesting that branded-content and licensing relationships can matter in media-oriented workflows. SM017
CM024 PixVerse's R1 API partner page targets studios, platform developers, and tool builders, making developers and API teams a distinct buyer class from end creators. SM016
CM025 Runway markets Characters as a real-time video-agent API for custom conversational characters, reinforcing the existence of an API-led developer segment in the category. SM008
CM026 TBRC says on-premise deployment offers more control and customization but requires more maintenance and upfront costs, which is a real enterprise tradeoff rather than just a technical option. SM002
CM027 PixVerse's English homepage claims enterprise-ready foundational models, APIs, and production-ready workflows, showing the company is courting enterprise rather than only consumers. SM013, SM014, SM015
CM028 Runway's pricing page reserves enterprise plans for teams scaling AI-video production with custom credit packages and contact sales, showing enterprise spend is sold differently from creator subscriptions. SM009
CM029 Budget ownership is fragmented across creator subscriptions, SMB campaign budgets, media production budgets, developer platform budgets, and enterprise transformation budgets. SM002, SM009, SM015, SM016, SM025
CM030 TBRC names expanding social media and digital platforms as a direct growth driver for the narrow market. SM002
CM031 Research and Markets and TBRC both point to automated editing, personalized marketing, real-time collaboration, and cloud-based collaborative video production as core trend or driver categories. SM001, SM002
CM032 PixVerse's English homepage says its world engine supports end-to-end multimodal generation, long-horizon streaming generation, and real-time 1080p video in interactive scenarios. SM014
CM033 Runway says it is building general world models and a real-time Characters API, showing that leading vendors are expanding toward interactive video agents, not only rendered clips. SM008
CM034 Pika markets agents, MCP integrations, and automation workflows, widening the category from pure generation into workflow tooling. SM024
CM035 PixVerse's pricing doc says $1 buys five V6 720p, 5-second, no-audio videos on the starter pack, while higher resolutions and audio consume more credits. SM012, SM013
CM036 Runway's pricing starts free, then scales through $12, $28, and $76 annualized monthly-equivalent plans before enterprise custom credits. SM009
CM037 Self-serve creator pricing and enterprise custom-credit plans let vendors land with low-friction experimentation and then expand into higher-value team workflows. SM009, SM012, SM015
CM038 China's 2023 generative AI measures apply to public services that generate text, images, audio, and video, and impose requirements around lawful data use, personal-information protection, content governance, transparency, labeling, risk mitigation, and filing or registration. SM018, SM019
CM039 China's 2025 AI-labeling measures require AI-generated online content to be labeled and are intended to curb false information and misuse. SM020, SM021
CM040 Compliance affects go-to-market speed because public video products in China must pair growth with labeling, governance, and filing operations rather than just creative feature velocity. SM018, SM019, SM020, SM021
CM041 BIS states that a license is required to export advanced computing items to China-linked entities even if those entities are located outside China, and GT Law says the requirement remained fully in force in 2026. SM022, SM023
CM042 Export controls and compute access remain adoption constraints for China-linked vendors because model iteration speed and cost still depend on advanced computing supply. SM022, SM023, SM005
CM043 OpenAI says the Sora web and app experiences were discontinued on April 26, 2026 and the Sora API is scheduled to end on September 24, 2026. SM010
CM044 Even category leaders can discontinue products, so buyers evaluating AI-video vendors still face platform-continuity and migration risk alongside pure model-quality questions. SM008, SM010, SM024
CM045 Public evidence does not isolate AIsphere's paid share, enterprise mix, or attributable spend capture inside the broader generative-video market, so a clean SAM or SOM remains unproven from public sources alone. SM001, SM002, SM015, SM016
CP001 PixVerse publicly positions itself as a full-stack AI media generation platform with APIs, enterprise-ready workflows, and service in more than 177 countries. SP001
CP002 Runway now frames itself as a world-model company with products such as GWM-1 and Characters, not only as an AI video generator. SP003, SP018
CP003 Runway publishes free, Standard, Pro, Max, and enterprise tiers and includes third-party models such as Kling, Seedance, and Veo in higher plans. SP004
CP004 OpenAI’s original Sora preview said the model could generate videos up to one minute long and framed Sora as progress toward real-world simulation. SP005
CP005 OpenAI’s deployed Sora experience was limited to up to 1080p, up to 20-second outputs and was publicly described as still struggling with unrealistic physics and long-duration actions. SP005, SP006
CP006 OpenAI later discontinued the Sora web and app experience on April 26, 2026 and plans to discontinue the Sora API on September 24, 2026. SP006, SP007
CP007 Kling 3.0 and Kling 3.0 Omni publicly emphasize deep multimodal instruction parsing, long-form storyboard control, native audio, and cross-scene consistency. SP008
CP008 Vidu markets text, image, and reference-video generation workflows for social content, ads, and storytelling. SP009
CP009 The fetched Hailuo landing page confirms a live AI video and image product but exposes little public detail on pricing, enterprise packaging, or governance. SP010
CP010 The fetched Wan landing page confirms a live AI video brand but leaves pricing and trust posture under-disclosed on the public surface. SP011
CP011 Jimeng publicly highlights text/image-to-video generation, first-and-last-frame control, Chinese prompt support, and a creator community. SP012
CP012 Pika has expanded into agent and MCP tooling and explicitly says agents can access “all the models,” making Pika as much a routing layer as a single-model product. SP013
CP013 Artificial Analysis compares quality, speed, and pricing across a broad video-model field that includes Kling, Hailuo, Vidu, Wan, Seedance, Sora, Grok, and open-weight LTX models. SP014
CP014 In Artificial Analysis’s retained text-to-video leaderboard, Seedance 2.0 720p leads at 1222 Elo, while Kling 3.0 1080p scores 1106, Vidu Q3 Pro 1082, PixVerse V6 1069, and Wan 2.6 1024. SP015
CP015 In Artificial Analysis’s retained image-to-video leaderboard, Seedance 2.0 720p leads at 1194 Elo, while Wan 2.7 scores 1093, PixVerse V6 1074, Kling 3.0 1080p 1072, and Vidu Q3 Pro 1061. SP016
CP016 PixVerse’s public platform docs say $1 buys five V6 720p five-second no-audio videos and show memberships scaling from $1,500 to $6,000 per month for larger plans. SP002
CP017 PixVerse’s public marketing site uses a vendor-authored chart to position PixVerse V6 against Grok, Kling, Veo, and Sora on quality, affordability, and speed. SP001
CP018 TechCrunch reported that Runway raised $315 million at a $5.3 billion valuation in February 2026 to expand its next generation of world models. SP018
CP019 TechCrunch reported that Moonshot AI raised about $2 billion at a $20 billion valuation and that its ARR topped $200 million in April 2026. SP019
CP020 a16z’s Top 100 Gen AI Apps report says the consumer AI ecosystem is stabilizing, app stores are cracking down on copycat apps, and Pixverse moved from the brink list into the core rankings. SP017
CP021 The same a16z report says many persistent AI consumer winners either use third-party or open models or operate as model aggregators, weakening the idea that model ownership alone wins distribution. SP017
CP022 a16z says Chinese video products such as Hailuo and Kling have exported globally and that Chinese video models had tended to outperform Western-developed models until Veo 3. SP017
CP023 China’s 2023 Interim Measures apply to public generative AI services for text, image, audio, video, and other content and require lawful data sources, transparency, stable service, and labeling through related rules. SP020, SP021
CP024 China’s March 2025 labeling guidance requires visible marks on AI-generated internet content and prohibits deleting, tampering with, fabricating, or concealing those labels. SP021, SP022
CP025 Loeb says China’s labeling measures extend explicit labels and metadata obligations to service providers and app distribution platforms, not only end users. SP022
CP026 Harris says complying with China’s AI labeling regime requires metadata-preserving workflows, agency governance, and platform-specific operational controls. SP023
CP027 NIST’s AI RMF and generative-AI profile signal that enterprise buyers increasingly expect structured documentation and trustworthiness practices around generative AI systems. SP024
CP028 BIS and Greenberg Traurig say advanced-computing exports to China-linked entities still require licenses even when the recipient sits outside China, preserving compute-supply diligence risk for China-linked labs. SP025, SP026
CP029 The EU AI Act requires identifiable AI-generated content and imposes documentation, copyright, and training-summary duties on providers of general-purpose AI models. SP027, SP028
CP030 Because the retained benchmark field already includes open-weight LTX models and Wan variants, internal build and self-hosted stacks remain credible substitutes for some technical buyers. SP014, SP015, SP016
CP031 Runway and Pika both teach buyers to think in multi-model terms, which lowers switching costs by letting the subscription or workflow layer route across changing frontier models. SP004, SP013
CP032 PixVerse’s most direct creator-facing peers are Runway, Sora, Kling, Vidu, Hailuo, Wan, Jimeng/Seedance, and Pika because each markets text-, image-, or audio-linked AI video creation rather than only assistant chat. SP001, SP003, SP005, SP008, SP009, SP010, SP011, SP012, SP013
CP033 The most powerful substitutes extend beyond standalone AI video apps to aggregators, internal build stacks, and adjacent AI workspaces that can satisfy the same content-creation job. SP004, SP013, SP014, SP017
CP034 Runway’s current scope now includes video agents and broader world-model products, so it competes for workflow ownership rather than just per-clip generation. SP003, SP018
CP035 Sora is a cautionary example that a headline entrant can preview frontier capability, launch with safety and quality caveats, and still retreat from the standalone product surface within the same cycle. SP005, SP006, SP007
CP036 Relative to Runway’s subscriptions, PixVerse’s official credit pricing looks competitively aggressive on short-form generation, but that does not yet prove realized enterprise pricing power. SP002, SP004
CP037 Independent leaderboard data shows PixVerse is first-tier but not clearly dominant because Seedance leads and Kling, Vidu, and Wan all remain close enough on quality and price to keep the category crowded. SP015, SP016
CP038 Competitive pressure in China is not limited to pure-play video startups because generalist labs like Moonshot can attract developers into broader coding, reasoning, and agent ecosystems instead of standalone video apps. SP017, SP019
CP039 Regulatory and trust posture may become a real competitive separator because China-facing vendors face visible labeling rules while EU- and U.S.-facing enterprise sellers face GPAI disclosure and AI-risk-management expectations. SP020, SP021, SP022, SP024, SP027, SP028
CP040 Switching costs at the model layer are low enough that any durable moat probably has to come from workflow ownership, contract stickiness, or proprietary data rather than raw model access alone. SP014, SP015, SP016, SP017, SP004, SP013
CP041 Public disclosure quality is uneven across the field, with Hailuo and Wan exposing substantially less retained detail on packaging and governance than Runway, OpenAI, or even PixVerse. SP010, SP011, SP003, SP006, SP001
CP042 The retained benchmark set shows a global and volatile frontier because Chinese leaders, Google’s Veo family, and xAI’s Grok all appear near the top at once. SP014, SP015, SP016
CP043 Because Runway bundles third-party models and Pika promises access to all models, buyers can multi-home without rewriting their workflow around one proprietary video brand. SP004, SP013
CP044 Publicly legible compliance can matter commercially because some consumer-first Chinese video apps expose limited governance material in the retained set while rulebooks keep expanding. SP010, SP011, SP021, SP022, SP023, SP027
CP045 AIsphere’s competitor set is bifurcating into destination apps, routing layers, and broader AI workspaces, which makes the category boundary porous and raises substitution risk from adjacent products. SP014, SP017, SP019
CI001 PixVerse maintains a public platform surface for AI video API access alongside its consumer creation surfaces. SI002, SI008
CI002 PixVerse’s official pricing docs state that $1 equals five V6 videos at 720p, five seconds, and no audio with the Starter pack. SI003
CI003 The same pricing docs meter usage by model, resolution, duration, audio, and motion mode rather than by a single unlimited flat rate. SI003
CI004 Official PixVerse pricing materials display prepaid credit packs from $50 to $5,000 and larger Scale and Business plan tiers, indicating both wallet top-ups and higher committed spend paths. SI003
CI005 PixVerse’s production-platform update says Team Plan pools credits, centralizes billing, and adds role-based permissions and shared asset libraries for organizations. SI004
CI006 PixVerse’s R1 partner program is selective rather than general availability and offers early access, grandfathered pricing, roadmap input, and technical support to qualified teams. SI005
CI007 PixVerse Canvas documentation says project costs are measured in credits that vary by model, inputs, and plan, while video generation is priced separately by model, length, and settings. SI006
CI008 PixVerse simultaneously maintains a consumer app, an app-download funnel, and developer or API surfaces, indicating a dual consumer-plus-platform distribution model. SI002, SI008, SI009, SI010
CI009 A16Z’s August 2025 top-100 consumer AI ranking says PixVerse moved from the prior mobile brink list into the core rankings, providing an external adoption proxy beyond company PR. SI011
CI010 CnTechPost, Yicai, and AI Insider all reported a $300 million March 2026 Series C round for AIsphere. SI012, SI013, SI014
CI011 Those March 2026 reports also repeated that AIsphere had surpassed $40 million in ARR and 100 million users, with more than 16 million monthly active users. SI012, SI013, SI014
CI012 KrASIA reported founder Wang Changhu’s statement that PixVerse subscription revenues already cover costs, but the claim is founder-reported rather than audited. SI016
CI013 PixVerse’s September 2025 PRNewswire release said the platform served more than 100 million users and had generated over 800 million videos. SI017
CI014 PixVerse’s 2026 production-platform update claimed 100M+ users across 175+ countries and more than 2 billion videos generated. SI004
CI015 The same production-platform update marketed about 68% cost reduction and about 57% faster production for organizations, which is evidence of enterprise value messaging rather than verified realized savings. SI004
CI016 The R1 partner program targets gaming studios, streaming platforms, and tool developers with required engineering capacity, launch timelines, and scale ambitions, implying a higher-touch enterprise sales motion. SI005
CI017 Team Plan features like shared workspaces, pooled credits, and administrator billing are consistent with a land-and-expand workspace model instead of pure one-seat creator subscriptions. SI004
CI018 Canvas supports task matrices, live queues, projected cost and time estimates, bulk export, and side-by-side model comparison, which fits professional workflow spend rather than casual one-off use. SI006
CI019 Google Play copy still markets PixVerse as a mass-market creation tool and points users toward the official hub and API integration, supporting a broad top-of-funnel consumer wedge. SI010
CI020 Longbridge’s Yicai summary said AIsphere is considering evolving its technology into an AI-native video game engine and cited interest from gaming and short-drama clients. SI015
CI021 Because official pricing varies with resolution, duration, audio, and motion mode, higher-fidelity or faster outputs likely consume more credits and expose customers to visibly higher unit prices. SI003
CI022 PixVerse’s updated R1 post says the product now supports continuous shared worlds, multi-user interaction, and real-time 1080p generation, features that likely raise serving complexity beyond single clip generation. SI007
CI023 Datadog’s 2025 10-K says cost of revenue includes payments to third-party cloud infrastructure providers for hosting software plus operations and support costs. SI019
CI024 Snowflake’s 2026 10-K warns that failing to meet minimum commitments under third-party cloud infrastructure agreements can negatively impact results of operations. SI020
CI025 Cloudflare’s 2025 10-K says cost of revenue includes co-location, network and bandwidth, certificate-authority, and equipment-depreciation expenses tied to serving paying customers. SI018
CI026 NVIDIA’s 2026 10-K says Blackwell Ultra increases token throughput and reduces cost per token versus Hopper, underscoring that AI-serving economics depend heavily on underlying infrastructure efficiency. SI021
CI027 Cloudflare disclosed $2,495.8 million of remaining performance obligations at year-end 2025 and expected to recognize 63% within 12 months, a transparency metric AIsphere does not publish. SI018
CI028 Snowflake reported 125% net revenue retention as of January 31, 2026 and growth in $1 million product-revenue customers from 576 to 733, while AIsphere discloses neither NRR nor large-account counts. SI020
CI029 Datadog reported $914.7 million of free cash flow in 2025 and explicitly monitors customers with ARR of $100,000 or more, another public quality marker absent from AIsphere’s disclosures. SI019
CI030 Adobe’s 2025 annual report disclosed $2.55 billion of cost of revenue, $12.51 billion of operating expenses, $7.13 billion of net income, and $10.03 billion of operating cash flow, highlighting the disclosure gap between mature creative-software leaders and AIsphere. SI022, SI023
CI031 Runway’s official pricing page uses the same basic design pattern as PixVerse—credits, tiered plans, and contact-sales enterprise packaging—suggesting AIsphere’s credit economics are not unusual for AI-video peers. SI024
CI032 TechCrunch reported that Runway raised $315 million at a $5.3 billion valuation in February 2026 to pre-train next-generation world models, showing that frontier AI-video platforms remain capital-hungry even at scale. SI025
CI033 Artificial Analysis publicly compares video-model quality, speed, and pricing across providers, which increases pricing transparency and competitive pressure in AI video. SI026
CI034 Public evidence proves that AIsphere has real monetization surfaces and list pricing, but it does not reveal realized customer pricing, discount bands, or segment revenue mix. SI002, SI003, SI004, SI005, SI006
CI035 The most plausible public revenue stack is a mix of consumer credits or subscriptions, team-workspace billing, API usage, and selective enterprise or partner contracts rather than a single homogeneous SaaS SKU. SI002, SI003, SI004, SI005, SI006
CI036 AIsphere’s public revenue-quality evidence is materially weaker than public software comparables because it does not disclose backlog or RPO, NRR, enterprise customer counts, renewal cohorts, or cash flow. SI018, SI019, SI020, SI022, SI023
CI037 Forward financing risk is lower than that of a pre-revenue startup because AIsphere closed a $300 million Series C and claims subscriptions cover costs, but the absence of cash, burn, runway, and debt data prevents a clean capital-adequacy conclusion. SI012, SI013, SI014, SI016
CI038 Capital intensity is likely highest in real-time, high-resolution, multi-user, and API-driven workloads rather than in lightweight template-led consumer clips. SI003, SI005, SI006, SI007
CI039 The company appears to be moving upmarket from creator acquisition into organizational sales by adding pooled billing, admin controls, API access, and partner programs. SI004, SI005, SI006
CI040 The repeated ARR and user claims raise confidence that AIsphere is commercially real, but their reliance on company-linked reporting keeps confidence in revenue quality at medium rather than high. SI012, SI013, SI014, SI016
CI041 The correct financial verdict is that AIsphere is financeable and monetizing, but still under-disclosed on margin, cash, contract quality, and concentration, so financial diligence should stay open. SI003, SI012, SI013, SI018, SI019, SI020
CE001 PixVerse’s public app shell now exposes Creation, Agent, Canvas, Mini-Apps, Marketing Hub, and API Platform in one surface. SE002
CE002 PixVerse’s homepage frames the company around frontier AI research, proprietary video foundation models, APIs, and a global video-intelligence ecosystem. SE001
CE003 PixVerse is delivering a multi-surface product stack across consumer app, creator web, workflow tools, and API surfaces rather than a single generation endpoint. SE001, SE002, SE005
CE004 PixVerse distributes official mobile access on iOS and Android and says the same account can be used across web and mobile with cloud sync. SE003, SE020
CE005 The 2026 production-platform update adds Team Plan shared workspaces, role-based permissions, a shared asset library, and pooled billing. SE008
CE006 PixVerse is introducing Mini Apps as purpose-built workflow shells above the core generation platform, with Ad Master as the first example. SE008
CE007 Canvas is described as a node-based workspace where references, scripts, storyboards, generation batches, and finished cuts live on one connected board. SE009
CE008 Canvas adds traceable prompts and settings, status tags, filtered views, batch queues, retry controls, and cost/time estimates for larger runs. SE009
CE009 PixVerse presents a real developer surface through a priced platform and company-authored claims of CLI/API integration into automated pipelines. SE005, SE008, SE012
CE010 PixVerse’s pricing docs show that generation is commercialized through credits across C1, V6, V5.6, audio, resolution, and duration settings. SE004
CE011 The V6 launch says PixVerse’s proprietary next-generation video model adds camera tracking, perspective shifts, depth-of-field control, multilingual text rendering, multi-shot audio, and CLI integration. SE006
CE012 The V6 review says V6 supports text-to-video, image-to-video, transition, extension, reference-to-video, 1–15 second duration, and up to 1080p output. SE006, SE012
CE013 PixVerse’s pricing docs and V6 review both state that 1080p V6 generation costs 18 credits per second without audio and 23 credits per second with audio. SE004, SE012
CE014 PixVerse’s public stack emphasizes integrated audio, multi-shot continuity, physics behavior, and literal camera control as part of the core product narrative. SE001, SE006, SE012
CE015 R1 is positioned as a real-time AI world model that generates continuous interactive 1080p video instead of isolated finished clips. SE007, SE010
CE016 PixVerse says R1 uses omni-native multimodal processing, 1–4 sampling steps per frame, and memory-augmented attention to support low-latency persistence. SE007
CE017 The R1 update adds single-photo avatars, removes the earlier five-minute session cap, and introduces multi-user shared worlds. SE007, SE010
CE018 The R1 partner page says the 720p real-time API provides synchronized audio and mid-stream narrative steering for interactive applications. SE011
CE019 PixVerse packages the R1 API as a selective partner program for gaming studios, streaming platforms, and tool developers rather than as broad self-serve general availability. SE011
CE020 PixVerse positions itself as a cinematic look-dev layer in front of Tripo Studio so teams can validate motion, lighting, and silhouette before committing to 3D modeling. SE013
CE021 KAGAMI Gate makes PixVerse the operating surface for a proof-of-concept that tracks and constrains licensed IP usage inside AI-generated video creation. SE014
CE022 The Captain Tsubasa rollout shows PixVerse translating that licensing layer into concrete user products: templates, challenges, and sanctioned fan-video creation. SE014, SE015
CE023 The Google Play listing markets a broad consumer feature set including text-to-video, image-to-video, transition, 4K upscale, extension, motion control, reference images, and speech integration. SE020
CE024 AppBrain reports about 73 million total PixVerse downloads, roughly 980 thousand downloads in the last 30 days, a 4.49 rating, and more than 4.3 million ratings on the access date. SE021
CE025 The archived Product Hunt page shows PixVerse with 304 followers, 6 reviews, and 4 launches, indicating some maker-community presence beyond company blogs. SE022
CE026 Artificial Analysis ranks PixVerse V6 14th in text-to-video with audio and V5.6 22nd on the accessed leaderboard. SE019
CE027 Artificial Analysis ranks PixVerse V6 10th in image-to-video with audio and V5.6 21st on the accessed leaderboard. SE018
CE028 PixVerse’s own V5.6 ranking post highlights selected leaderboard views, price framing, and “top-tier” language rather than claiming blanket category leadership. SE017, SE018, SE019
CE029 PixVerse’s public marketing emphasizes high Elo and favorable price positioning, but the independent with-audio leaderboards still place V6 in the first tier rather than at number one. SE001, SE018, SE019
CE030 PixVerse’s own V6 review says the model is strongest on camera movement, continuity, short narrative structure, and audio-visual timing, but still needs retries for chaotic action, multilingual dialogue, and brand-accurate product shots. SE012
CE031 AppBrain user comments include complaints about unwanted changes, unnatural renders, and wasted credits even alongside praise for free credits and newer voice-model options. SE021
CE032 Across the reviewed official home, app, blog, and platform pages, no public trust center, SOC 2 page, ISO 27001 page, or public uptime/status page was visible. SE001, SE002, SE005, SE008
CE033 China’s 2023 generative-AI measures apply to public services that generate text, image, audio, video, or other content in China and require lawful data handling, privacy protection, output labeling, and safe stable service. SE023
CE034 China’s 2025 labeling measures require visible labels and metadata-style identifiers for AI-generated content, and distribution platforms must review whether apps providing generative AI comply. SE024, SE025
CE035 Harris Sliwoski argues that operational compliance now requires workflow rules, metadata retention, partner governance, and fast takedown processes because platform enforcement is tightening. SE028
CE036 BIS guidance in May 2026 confirmed that advanced-computing export licenses remain required for China-linked entities even when the recipient is physically outside China. SE026, SE027
CE037 For a China-linked AI-video platform, product delivery depends not only on model quality but also on compliant access to advanced compute and auditable content-governance workflows. SE023, SE024, SE026, SE027
CE038 PixVerse’s AI for Good partnership and workshop/film-festival program show the company is trying to frame its product as globally useful creator infrastructure rather than only a consumer effect app. SE016
CE039 PixVerse’s public materials repeat that the platform serves 100 million-plus creators and enterprises across 177-plus countries, but that scale is still company-claimed rather than independently audited product telemetry. SE016, SE017
CE040 Even when creation starts on mobile, PixVerse remains cloud-served and credit-metered rather than on-device inference, so user experience still depends on backend reliability and spend tolerance. SE003, SE020, SE021
CE041 The public product stack can be modeled as four layers: consumer creation surfaces, collaborative workflow surfaces, programmable API surfaces, and an emerging world-model surface. SE002, SE008, SE009, SE011
CE042 PixVerse mixes proprietary models such as V6, C1, and R1 with selective external-model access, making workflow breadth a more durable differentiator than any single model release alone. SE006, SE011, SE013, SE020
CE043 The selective R1 partner program can improve roadmap fit for high-scale builders, but it also means the newest real-time surface is not yet broadly proven in self-serve production. SE011
CE044 PixVerse’s workflow ownership, IP-licensing experiments, and partner integrations differentiate the product more clearly than raw leaderboard placement does. SE009, SE013, SE014, SE018, SE019
CE045 Public enterprise-readiness evidence is still thinner than product breadth because reviewed surfaces do not disclose API SLA terms, uptime history, incident reporting, or moderation/provenance implementation depth. SE005, SE008, SE011, SE012
CE046 OpenAI publicly discloses C2PA metadata, visible watermarks, and deepfake mitigations for Sora, providing a benchmark that PixVerse’s reviewed public safety surfaces do not yet visibly match. SE029, SE001, SE002
CE047 Apptopia lists PixVerse: AI Video Generator on iOS under MOTIVAI PRIVATE LIMITED in Photo & Video and Entertainment, corroborating active iOS distribution beyond company-authored pages. SE030
CU001 PixVerse maintains consumer app, app-download, and platform/API surfaces, implying distinct self-serve creator and developer entry points. SU001, SU002, SU009
CU002 Google Play, Uptodown, APKPure, and AppBrain all distribute or track the Android app, while Apptopia tracks iOS, showing multi-store distribution rather than a single-channel mobile presence. SU003, SU004, SU005, SU006, SU007
CU003 Google Play positions PixVerse as an all-in-one AI video maker for creators and highlights text-to-video, image-to-video, upscale, extension, and selection editing features. SU003
CU004 Uptodown describes PixVerse as beginner-friendly for transforming images or text prompts into animated video with trending effects. SU006
CU005 APKPure explicitly says PixVerse suits creators, marketers, and anyone needing social-friendly videos from prompts or photos. SU007
CU006 PixVerse’s app-download page emphasizes mobile creation, image animation, social clips, and fast generation for creators anywhere. SU002
CU007 PixVerse’s production-platform update says professional creators and enterprises pushed the product toward team workflows, asset management, automation, and developer tools. SU011
CU008 The Team Plan adds shared workspaces, role-based permissions, shared asset libraries, and consolidated billing with pooled credits. SU011
CU009 PixVerse’s platform site and pricing docs show a separate API or developer-platform surface alongside its consumer creation surfaces. SU009, SU010
CU010 PixVerse’s R1 partner program is aimed at studios, platform developers, and tool builders who want to integrate real-time AI video generation into their products. SU012
CU011 The R1 partner program is selective rather than general availability and offers qualified teams early access, favorable pricing, and roadmap influence. SU012
CU012 AppBrain’s PixVerse page displays both 50,000,000+ downloads and a separate 73 million lifetime-download estimate, evidencing very large Android reach even if the exact methodology is opaque. SU004
CU013 AppBrain says PixVerse drew roughly 980 thousand downloads in the last 30 days and ranked #2 in photography at the time of capture. SU004
CU014 AppBrain rates PixVerse 4.49 out of 5 based on roughly 4.3 million ratings. SU004
CU015 The archived Product Hunt page shows 304 followers, six reviews, four launches, and a 4.3 score, indicating some prosumer community interest but a much smaller footprint than the mass-market app stores. SU008
CU016 Apptopia shows a PixVerse iOS App Store listing under MOTIVAI PRIVATE LIMITED in the Photo & Video and Entertainment categories. SU005
CU017 PixVerse’s September 2025 PRNewswire release said the platform served more than 100 million users worldwide and had generated over 800 million videos. SU023
CU018 CnTechPost, Yicai, and AI Insider all reported in March 2026 that AIsphere had surpassed 100 million users and more than $40 million ARR. SU024, SU025, SU026
CU019 Yicai and AI Insider also cited more than 16 million monthly active users, while official 2026 company materials claimed reach across 175+ countries and more than 2 billion videos generated. SU011, SU025, SU026
CU020 A16Z’s August 2025 consumer AI ranking said PixVerse moved from the prior mobile Brink List into the core rankings, giving an external adoption proxy beyond company PR. SU027
CU021 PixVerse’s R1 invite guide says paid subscribers automatically receive R1 access and lists Plus ($10/month), Pro ($25/month), and Team ($58/month) plans. SU014
CU022 The same invite guide says PixVerse distributes invite codes through Twitter/X, Discord, YouTube, TikTok, and Instagram, with Discord described as the most active channel. SU014
CU023 PixVerse’s quality guide says R1 supports 360p, 480p, 720p, and 1080p but defaults to lower resolution for smoother real-time streaming. SU015
CU024 The same guide recommends 1080p for final output, implying a prosumer or paid workflow beyond casual preview usage. SU015
CU025 PixVerse’s Dancing Baby page says the trend has gained millions of views across social media and frames template use as a one-click workflow for everyday creators. SU016
CU026 The Winter Sovereign and Fly to the Sun pages show that PixVerse’s template library spans fantasy and cinematic storytelling, not just one meme format. SU017, SU018
CU027 PixVerse’s Captain Tsubasa collaboration lets users generate videos with licensed franchise characters through official templates until July 26, 2026. SU019
CU028 The KAGAMI Gate proof-of-concept positions PixVerse as the AI video platform inside a system for licensing and tracking IP use in generative video. SU020
CU029 Taken together, the Captain Tsubasa and KAGAMI Gate pages show PixVerse pursuing IP/licensing use cases rather than only generic creator tooling. SU019, SU020
CU030 The Tripo Studio integration positions PixVerse as a rapid look-development stage for game teams before they convert a winning frame into a 3D asset. SU021
CU031 The UN AI for Good partnership invites creators worldwide to submit AI video works for a film festival and ties PixVerse to a UN-stage institutional program. SU022
CU032 Google Play also surfaces the Captain Tsubasa collaboration directly in the app-store listing, showing that licensed campaigns are used for acquisition as well as for in-product engagement. SU003, SU019
CU033 AppBrain explicitly notes that some competitors may offer a larger template or motion library and that broad permissions or SDK integrations may raise privacy concerns for sensitive users. SU004
CU034 Official R1 and Team Plan materials prove monetization paths and professional workflows, but they do not name specific paying enterprise accounts or contract values. SU011, SU012, SU014, SU015
CU035 No reviewed public source discloses PixVerse’s paying-customer count, paying-team count, or free-to-paid conversion by segment. SU003, SU004, SU011, SU014, SU023, SU024, SU025, SU026, SU028
CU036 No reviewed public source discloses NRR, GRR, customer churn, or logo-retention cohorts for any PixVerse segment. SU011, SU012, SU014, SU023, SU024, SU025, SU026, SU028
CU037 No reviewed public source discloses contract length, renewal timing, or renewal rates for Team Plan, API partners, or named partner programs. SU011, SU012, SU019, SU020, SU022
CU038 Strong consumer ratings are directionally useful, but ratings and reviews are not evidence of enterprise retention, expansion, or revenue durability. SU004, SU008
CU039 No reviewed public source discloses customer concentration, top-account share, or revenue contribution from named partner programs. SU011, SU012, SU019, SU020, SU021, SU022, SU023, SU024, SU025, SU026, SU028
CU040 No reviewed public source discloses the revenue mix between consumer subscriptions or credits and enterprise, API, or partner contracts. SU010, SU011, SU012, SU023, SU024, SU025, SU026, SU028
CU041 PixVerse’s expansion logic runs from self-serve creation and template virality into paid R1 access, team collaboration, and API partnerships. SU002, SU011, SU012, SU014, SU016, SU017, SU018
CU042 The named public proof set is skewed toward partner and use-case proof—Captain Tsubasa, KAGAMI Gate, UN AI for Good, and Tripo Studio—rather than toward named enterprise buyers with contract or outcome disclosure. SU019, SU020, SU021, SU022
CU043 Channel dependence is material because discovery and usage are visibly mediated by app stores, social invite channels, Product Hunt, and partner ecosystems rather than by a disclosed direct-sales base. SU003, SU005, SU008, SU014, SU021
CU044 KrASIA reported founder commentary that subscriptions cover costs, but even that financing story did not disclose customer mix, retention, or concentration. SU028
CU045 Production-platform and Canvas pages describe project, campaign, and client organization, which is more consistent with agency or team workflows than with hobby-only usage. SU011, SU013
CU046 PixVerse’s V6 review frames the product as a production tool with repeatable workflows and concrete credit-cost testing, reinforcing appeal to prosumers and team users rather than only hobbyists. SU029
CR001 China's interim generative-AI rules apply to public services in the PRC that generate text, images, audio, video, or other content. SR015, SR016
CR002 Those rules require lawful training-data sources and prohibit infringement of third-party intellectual-property rights during training and optimization. SR015, SR016
CR003 China's generative-AI regime links providers to labeling obligations for generated images and video through the deep-synthesis and service rules. SR015, SR017, SR018, SR019
CR004 China's AI-labeling measures take effect on 2025-09-01 and require both visible labels and metadata-oriented labeling methods. SR017, SR018, SR019
CR005 Providers also bear user-agreement, complaint-handling, security, and personal-information obligations under the PRC service rules. SR015, SR016
CR006 Violations of China's generative-AI measures can trigger warnings, orders to correct, or suspension of related services. SR015, SR016
CR007 BIS clarified in May 2026 that advanced-computing items still require export licenses for China-linked entities despite any temporary enforcement pause. SR020, SR021
CR008 NVIDIA warns that export-control rules can restrict product availability and add cost and operational complexity to advanced-computing supply. SR029
CR009 Generative-AI copyright litigation in 2026 still leaves open questions on training-data infringement, fair use, and ownership of AI outputs. SR026, SR027, SR028
CR010 For an AI-video company, copyright exposure can arise from both training-data provenance and the licensed or infringing character of generated outputs. SR026, SR027, SR028, SR034
CR011 PixVerse is offered as a public web and mobile product with Chinese and international surfaces rather than a private internal-only tool. SR001, SR008, SR009
CR012 PixVerse also markets APIs and production workflows, which expands the compliance surface beyond consumer entertainment into enterprise use cases. SR002, SR003, SR005, SR006
CR013 R1 is marketed as a real-time world model that responds continuously to user input, which raises live-content moderation and labeling complexity beyond static clip generation. SR004, SR015, SR017
CR014 The EU AI Act creates transparency and documentation obligations around general-purpose AI models and downstream use that can matter to API and enterprise workflows. SR022, SR023, SR024
CR015 NIST's AI Risk Management Framework treats governance, monitoring, testing, and incident response as ongoing controls rather than one-time certifications. SR025
CR016 Public evidence does not show a published PixVerse compliance audit pack, DPA, or detailed China and EU implementation artifacts. SR001, SR002, SR005
CR017 AIsphere has raised a $300 million Series C, but public burn, runway, and audited profitability remain undisclosed. SR013, SR001
CR018 PixVerse's pricing docs show low-cost credit-based video generation, making unit economics sensitive to model mix, duration, resolution, and audio settings. SR003
CR019 Artificial Analysis places PixVerse among leading video-model vendors, proving relevance but raising expectation risk if rankings slip. SR007
CR020 Runway still markets world-model ambitions, illustrating that PixVerse competes against well-funded rivals with similar strategic narratives. SR012, SR007
CR021 OpenAI's Sora discontinuation shows that AI-video product packaging and category leadership can change abruptly even among major incumbents. SR011
CR022 AppBrain shows PixVerse at more than 50 million downloads and over 4 million reviews, proving scale but also dependence on app-store discovery and payments. SR009
CR023 APKPure describes PixVerse as a fast consumer tool for creators and marketers, reinforcing a broad consumer-distribution orientation rather than a narrow enterprise-only model. SR010
CR024 Sina reported that Alibaba Cloud provides AIsphere with full-stack AI support for training, inference, global deployment, and security-compliance capabilities. SR014
CR025 Alibaba Cloud dependence means a cloud-pricing or capacity change can propagate into service availability, latency, and gross margins. SR014, SR030, SR032
CR026 A selective R1 API partner program suggests enterprise monetization is being curated rather than broadly self-serve, which can slow scale and concentrate onboarding risk. SR006, SR002
CR027 KAGAMI Gate is only a proof-of-concept for managing licensed IP in AI video, so AIsphere's rights-management mitigation is promising but immature. SR033, SR034
CR028 The Captain Tsubasa collaboration shows branded IP can expand growth while also tying campaigns to external licensors and renewal risk. SR034, SR033
CR029 The Tripo Studio integration broadens creator workflows but adds ecosystem dependence on external partners to complete the value proposition. SR035, SR005
CR030 Public sources do not disclose top-customer concentration, NRR, or retention cohorts, leaving customer durability unresolved despite strong app-scale evidence. SR013, SR009
CR031 Cloudflare, Datadog, and Snowflake each flag infrastructure outages, security incidents, and service interruptions as material platform risks in their 2025 and 2026 filings. SR030, SR031, SR032
CR032 AIsphere's public materials emphasize rapid product expansion, APIs, and team workflows, which increases operational complexity even while broadening monetization avenues. SR004, SR005, SR006
CR033 Real-time video features expand the surface for moderation misses, mislabeled outputs, and response-time failures during traffic spikes. SR004, SR017, SR025
CR034 Public evidence shows product breadth and growth claims, but not a verified incident history or enterprise SLA schedule. SR002, SR005, SR009
CR035 NVIDIA supply remains strategically critical because frontier video models depend on advanced computing and export rules can affect China-linked access. SR020, SR021, SR029
CR036 AIsphere's combination of public consumer scale, API ambitions, and China domicile makes legal and regulatory risk more material than for a private internal AI lab. SR001, SR008, SR015, SR017
CR037 Pricing compression risk is heightened because PixVerse competes in a crowded AI-video field where leaders market rapidly improving capabilities. SR003, SR007, SR012
CR038 The $300 million round reduces near-term financing risk but does not remove capital risk while burn, cloud commitments, and customer payback are undisclosed. SR013, SR014
CR039 Public evidence does not reveal the depth of compliance staffing, solutions engineering, or leadership redundancy needed for global enterprise expansion. SR001, SR005, SR006
CR040 Because AIsphere markets both Chinese and global products, a regulatory incident in China could spill into foreign distribution, enterprise sales, and valuation perception. SR015, SR017, SR020
CR041 The top three ranked risks are China compliance and labeling, GPU and cloud concentration, and unresolved copyright or IP exposure. SR015, SR020, SR026
CR042 Model-quality and consumer-distribution risk sit just below the top tier because app-store traction and benchmark leadership can reverse quickly in AI video. SR007, SR009, SR012
CR043 Capital and execution risk remain mid-high rather than critical because financing is strong but operational and financial disclosures are incomplete. SR013, SR014, SR005
CR044 Visible mitigations include licensed-IP experiments, a selective partner program, production-workflow tools, and governance frameworks rather than any single regulatory moat. SR005, SR006, SR025, SR033, SR034
CR045 The clearest thesis-break events are a China compliance action, inability to secure frontier GPU capacity, or sustained quality and distribution slippage that blocks enterprise conversion. SR015, SR020, SR007, SR009
CR046 Regulatory friction and infrastructure constraints can transmit into enterprise trust, revenue conversion, margin quality, and ultimately valuation in the same downside pathway. SR015, SR020, SR014, SR013
CR047 The dependency map should center on advanced compute, Alibaba Cloud, app stores, API partners, rights licensors, and creator-tool integrations as the main external nodes. SR014, SR006, SR033, SR035
CR048 The most useful monitoring indicators are CAC and labeling artifacts, GPU-capacity lead times, app-store ratings and downloads, benchmark placement, partner conversions, and disclosed runway. SR017, SR020, SR009, SR007, SR006, SR013
CR049 LexisCN's translated text confirms the 2023 generative-AI rules were jointly issued by seven PRC ministries and took effect on 2023-08-15. SR036
CR050 Regulations.AI's summary of the 2025 labeling measures emphasizes both visible labels and machine-readable metadata or watermark-style identifiers for AI-generated content. SR037
CR051 Comparative AI characterizes the PRC generative-AI interim measures as a departmental rule rather than a higher-level State Council or NPC instrument, underscoring that future policy tightening remains possible. SR038
CV001 Multiple March 2026 reports confirm that AIsphere raised a $300 million Series C and frame it as a record financing event for China's AI-video segment. SV001, SV002, SV003, SV029
CV002 The retained public record supports only a directional unicorn-level AIsphere mark around $1 billion rather than an exact reconciled post-money valuation. SV001, SV002, SV003, SV029
CV003 March 2026 coverage repeated that AIsphere had surpassed $40 million in ARR, more than 100 million users, and more than 16 million monthly active users. SV001, SV002, SV003
CV004 KrASIA reported founder Wang Changhu's statement that PixVerse subscription revenue already covers costs, but the claim is founder-reported rather than audited. SV004
CV005 Official product surfaces show PixVerse sells through self-serve credits, larger plans, and enterprise-style packaging rather than through a single free consumer channel. SV006, SV007
CV006 The retained public sources do not disclose AIsphere's exact post-money ownership, liquidation preferences, anti-dilution terms, or side-letter economics. SV001, SV002, SV003, SV004
CV007 If AIsphere's headline mark was roughly $1 billion and ARR merely exceeded $40 million, the implied headline multiple is about 25x ARR or lower because the disclosed ARR figure is only a floor. SV001, SV002, SV003
CV008 AIsphere's implied ~25x ARR ceiling is below the roughly ~59x revenue multiple implied by Runway's $5.3 billion valuation on Sacra's $90 million annualized-revenue estimate. SV012, SV013, SV014
CV009 Runway's February 2026 financing re-rated the company to $5.3 billion while the product expanded from AI video into world-model and API workflows. SV012, SV013, SV014, SV030
CV010 Runway is a closer business-model comp than broad AI labs because official and third-party sources show both self-serve video plans and enterprise/API monetization. SV011, SV013, SV030
CV011 Sacra and FirmKnow both place Pika's 2024 valuation at about $470 million after its $80 million Series B, providing a lower-end private AI-video comparison point. SV016, SV017
CV012 Pika's public record still lacks a clean audited revenue base, so the company is useful for valuation bracket context but not for a reliable multiple comparison. SV015, SV016, SV017
CV013 TechCrunch and The AI Rankings say Moonshot raised about $2 billion at about a $20 billion valuation in May 2026, and TechCrunch says ARR topped $200 million in April 2026. SV021, SV022, SV023
CV014 Moonshot's reported valuation implies roughly a 100x ARR multiple on the disclosed >$200 million floor, an upper-bound valuation profile far above AIsphere's surface multiple. SV021, SV022, SV023
CV015 Business Day and WinBuzzer put MiniMax around a $6.5 billion IPO valuation in early 2026, while WinBuzzer says it had only $53 million of first-nine-month 2025 revenue and a $512 million loss. SV019, SV020
CV016 Using WinBuzzer's revenue figure, MiniMax floated at roughly 90x-120x sales depending on whether investors annualize the nine-month run rate, showing how narrative-heavy public AI valuations had become. SV019, SV020
CV017 Startup Wired and Business Day say Zhipu targeted roughly $640 million of IPO proceeds at about a $6.7 billion or HK$51.2 billion valuation. SV018, SV019
CV018 TechCrunch later reported that Zhipu traded at roughly a $55.9 billion market cap and MiniMax at roughly $33 billion after model-driven rallies, showing that debut pricing can materially understate immediate speculative appetite. SV022
CV019 Artificial Analysis leaderboards place PixVerse V6 in the competitive first tier but behind Seedance on text-to-video and inside a tight image-to-video pack with Wan, Kling, and Vidu. SV008, SV009, SV010
CV020 Because the benchmark gaps are narrow and the category is crowded, AIsphere lacks clear evidence of a product lead strong enough to warrant Moonshot-like or even Runway-like premium multiples on quality alone. SV008, SV009, SV010, SV011, SV015, SV030
CV021 Adobe's 2025 filing discloses cost of revenue, operating expenses, net income, and operating cash flow while AIsphere provides no comparable public revenue-quality or cash-flow ledger. SV024
CV022 NVIDIA's 2026 filing emphasizes token-throughput and lower cost per token as core AI economics, underscoring that AI-video value creation depends on compute efficiency that AIsphere does not quantify publicly. SV025
CV023 Snowflake's 2026 filing discloses 125% net revenue retention and growth in $1 million product-revenue customers, metrics absent from AIsphere's public record. SV026
CV024 Datadog's 2025 filing discloses free cash flow and the count of customers with ARR above $100,000, again highlighting the quality markers AIsphere does not publish. SV027
CV025 Cloudflare's 2025 filing details hosting, bandwidth, and support-cost drivers inside cost of revenue, giving a public template for economics disclosure that AIsphere lacks. SV028
CV026 AIsphere's surface multiple looks lower than hot AI peers, but that discount mostly compensates for weaker disclosure, smaller scope, and less proven monetization rather than creating an obvious bargain. SV001, SV003, SV012, SV013, SV020, SV021, SV022, SV024, SV026
CV027 Business 2.0 News explicitly argues that high AI valuations face downward pressure if developer traction fails to convert into enterprise revenue within 12 to 18 months; that logic is directly relevant to AIsphere. SV023
CV028 MiniMax's IPO article shows that investors tolerated large losses despite litigation and export-control risk, which proves public appetite existed but also shows how narrative-sensitive AI valuations can be. SV020
CV029 AIsphere has enough real scale signals—major funding, public pricing, global user claims, and independent benchmark presence—to stay investable rather than avoidable. SV001, SV003, SV005, SV006, SV007, SV008, SV009, SV010
CV030 Absent cap-table and revenue-quality disclosure, the supportable current recommendation is research-more rather than buy. SV001, SV002, SV004, SV024, SV026
CV031 Confidence should remain medium because the valuation direction is plausible, but exact pricing mechanics, ARR quality, and security terms remain under-documented. SV001, SV002, SV003, SV004, SV024, SV026
CV032 Risk should remain high because valuation support depends simultaneously on revenue conversion, compute discipline, and a favorable AI capital-markets window. SV012, SV020, SV022, SV023, SV025
CV033 The valuation stance is stretched rather than outright expensive because the implied ~25x ARR ceiling is below hotter comps, but the underlying ARR floor and term sheet are not strong enough to create a clean margin of safety. SV001, SV002, SV003, SV012, SV013, SV021, SV022
CV034 In a bull case, audited recurring revenue quality, cleaner enterprise/API mix, and sustained first-tier product proof could support a roughly $1.4 billion to $1.8 billion valuation band. SV007, SV008, SV009, SV010, SV012, SV013
CV035 In a base case, a roughly $0.85 billion to $1.1 billion band is supportable if ARR proves real but revenue quality and terms stay only partly disclosed. SV001, SV002, SV003, SV012, SV013, SV016, SV017
CV036 In a bear case, a roughly $0.55 billion to $0.75 billion band is plausible if paid conversion, enterprise quality, or sector scarcity premiums disappoint. SV015, SV020, SV023
CV037 A higher AIsphere valuation than the current implied mark would require an audited revenue bridge, better enterprise/API mix disclosure, and cleaner security terms more than another headline funding round. SV004, SV024, SV025, SV026, SV027, SV028
CV038 A lower entry price would be justified if ARR quality looks weaker than the >$40 million headline, if enterprise mix is thin, or if preference terms materially subordinate new money. SV001, SV002, SV004, SV023
CV039 The cleanest bull signal would be evidence that paid ARR is closer to Runway-like software revenue than to creator-heavy promotional or subsidized usage. SV013, SV016, SV017, SV024, SV026
CV040 A next round below the implied unicorn mark or with heavy senior preferences would show that the headline valuation overstated external support or overstated security quality. SV001, SV002, SV003
CV041 Evidence that active paid customers, renewals, or API usage lag the user headline would break the thesis that AIsphere's scale signals deserve premium economics. SV003, SV005, SV007, SV024, SV026, SV027
CV042 If Hong Kong or private-market appetite for Chinese AI cools materially, AIsphere could lose external scarcity support even if company execution remains adequate. SV019, SV020, SV022, SV023
CV043 The most important diligence ask is a monthly bridge from bookings to ARR to recognized revenue by consumer, API, and enterprise channels. SV001, SV003, SV006, SV007, SV024, SV026, SV027
CV044 The second critical diligence ask is the March 2026 cap table, liquidation waterfall, anti-dilution terms, and any investor side letters. SV001, SV002, SV003
CV045 The third critical diligence ask is unit-economics disclosure: compute cost per video minute, gross margin by product surface, and cloud commitments. SV007, SV025, SV028
CV046 The fourth critical diligence ask is cohort evidence such as paid customer count, net revenue retention, churn, and concentration because public comps all disclose analogous quality markers. SV024, SV026, SV027
CV047 Taken together, the retained evidence supports AIsphere as a credible but still under-documented AI-video asset whose company quality may be better than current security visibility. SV001, SV006, SV008, SV012, SV021, SV024
来源
编号出版方标题引文
SO001 AIsphere 爱诗科技AIsphere 爱诗科技致力于打造全球领先的AI视频生成大模型及应用。
SO002 PixVerse Frontier AI Research and Products that Redefine the Future of Video Intelligence Enterprise-ready AI video foundational models and solutions, delivering real business impact.
SO003 PixVerse PixVerse | Create Amazing AI Videos from Text & Photos with AI Video Generator Transform your photos into captivating AI videos with PixVerse's powerful AI model.
SO004 PixVerse Blog Product Updates
SO005 PixVerse Blog PixVerse Launches V6: Advancing AI Video Generation With Cinematic Precision PixVerse has launched V6, the next generation of its proprietary AI video model.
SO006 PixVerse Blog PixVerse Launches R1: The First Real-Time AI World Model R1 generates continuous, interactive 1080p video that responds to user input in real time.
SO007 PixVerse PixVerse R1 Explained: Real-Time AI Video World Model PixVerse R1 is a real-time AI world model for interactive video generation.
SO008 PixVerse PixVerse V5.6 Ranks #2 on Artificial Analysis Leaderboard PixVerse V5.6 has claimed the #2 position worldwide.
SO009 CnTechPost AI video startup AIsphere raises $300 million in record China funding The funding round was led by CDH Investments, with more than 20 institutions participating in the capital injection.
SO010 Yicai Global via r.jina.ai AIsphere Raises USD300 Million, Most by a Chinese Text-to-Video Startup, Report Says The company has more than USD40 million in annual recurring revenue at the end of last year.
SO011 Longbridge / Yicai AIsphere Raises USD300 Million, Most by a Chinese Text-to-Video Startup, Report Says
SO012 The AI Insider AIsphere Raises $300M to Scale PixVerse AI Video Platform and Global Expansion
SO013 KrASIA AIsphere touts PixVerse as the “Canva for video generation,” and lands the funding to prove it AIsphere’s flagship product, PixVerse, has surpassed 100 million global users, up from 60 million just months ago.
SO014 South China Morning Post A promising Chinese start-up rival to OpenAI’s Sora raises US$14 million
SO015 Yicai Global via Wayback PixVerse Owner AISphere Bags USD55.2 Million in Series A Fundraisers PixVerse boasts more than 40 million users worldwide.
SO016 PR Newswire PixVerse Raised $60M Series B to Accelerate Global AI Video Adoption PixVerse now serves more than 100 million users worldwide, who have collectively produced over 800 million videos.
SO017 TMTPOST Alibaba Leads $60 Million Series B Round in AI Video Startup AISphere Founded in April 2023, AISphere develops world-leading AI video generation models and applications.
SO018 EqualOcean Aishi Technology Launches PixVerse R1, a Real-Time World Model Supporting 1080P Resolution Aishi Technology officially released PixVerse R1, the world’s first general-purpose real-time world model supporting resolutions of up to 1080P.
SO019 Sina Finance 爱诗科技与阿里云达成全栈AI合作 双方将围绕AI全栈能力与全球化布局,在模型、算力、产品、生态和商业层面建立深度协同。
SO020 Baidu Baike 北京爱诗科技有限公司 北京爱诗科技有限公司于2023年04月07日成立。
SO021 Baidu Baike 王长虎 现任爱诗科技有限公司创始人,曾任字节跳动视觉技术负责人。
SO022 Andreessen Horowitz The Top 100 Gen AI Consumer Apps – 5th Edition
SO023 Google Play PixVerse: AI Video Generator - Apps on Google Play PixVerse is your all-in-one tool for creating stunning videos.
SO024 PixVerse PixVerse App Download: Get the AI Video Generator App for Mobile Text-to-Video & Image Animation
SO025 PixVerse Blog Download PixVerse App For Free: AI Video Generator on iOS & Android The PixVerse mobile app brings the full power of the world’s leading AI video generator right to your smartphone.
SM001 Research and Markets Generative AI in Video Creation Market Report 2026
SM002 The Business Research Company Generative AI In Video Creation Market Report 2026
SM003 Fortune Business Insights Generative AI Market Size, Share, Value Report [2026-2034]
SM004 Andreessen Horowitz The Top 100 Gen AI Consumer Apps – 4th Edition
SM005 Artificial Analysis Video Model Comparisons
SM006 Artificial Analysis Text to Video Leaderboard - Top AI Video Models
SM007 Artificial Analysis Image to Video Leaderboard - Top AI Video Models
SM008 Runway Runway | Building AI to Simulate the World
SM009 Runway AI Image and Video Pricing from $12/month | Runway AI
SM010 OpenAI What to know about the Sora discontinuation | OpenAI Help Center
SM011 Google Play PixVerse: AI Video Generator - Apps on Google Play
SM012 PixVerse Pricing - PixVerse Platform Docs
SM013 PixVerse PixVerse Platform
SM014 PixVerse PixVerse English homepage
SM015 PixVerse PixVerse Evolves From Creation Tool to Full Production Platform
SM016 PixVerse PixVerse R1: 720p Real-Time Video API and Partner Program Now Open
SM017 PixVerse PixVerse blog home
SM018 regulations.ai Interim Measures for the Administration of Generative AI Services
SM019 China Law Translate Interim Measures for the Management of Generative Artificial Intelligence Services
SM020 State Council Information Office / Xinhua China requires labeling of AI-generated online content
SM021 Harris Sliwoski China’s New AI Labeling Rules: What Every China Business Needs to Know
SM022 Bureau of Industry and Security Homepage | Bureau of Industry and Security
SM023 Greenberg Traurig BIS Clarifies Ongoing License Requirement for Advanced Computing Items to China-Linked Entities
SM024 Pika Pika homepage
SM025 Vidu AI Video Generator for Text, Image & Reference Videos
SM026 Kling AI Kling AI: Next-Generation AI Creative Studio
SM027 Jimeng AI 即梦AI - 即刻造梦
SM028 Wan AI Wan AI: Leading AI Video Generation Model
SP001 PixVerse Frontier AI Research and Products that Redefine the Future of Video Intelligence
SP002 PixVerse Platform Docs Pricing - PixVerse Platform Docs $1 = 5 videos (v6, 720p, 5s, no audio, with Starter pack).
SP003 Runway Runway | Building AI to Simulate the World
SP004 Runway AI Image and Video Pricing from $12/month | Runway AI All AI images and video models (Gen-4.5, Nano Banana Pro, Aleph, Veo 3.1, and more).
SP005 OpenAI Sora: Creating video from text Sora can generate videos up to a minute long while maintaining visual quality and adherence to the user’s prompt.
SP006 OpenAI Sora is here The version of Sora we are deploying has many limitations. It often generates unrealistic physics and struggles with complex actions over long durations.
SP007 OpenAI Help Center What to know about the Sora discontinuation The Sora web and app experiences were discontinued on April 26, 2026.
SP008 Kling AI Kling AI: Next-Generation AI Creative Studio VIDEO 3.0 and VIDEO 3.0 Omni natively support deep multimodal instruction parsing and cross-task integration.
SP009 Vidu AI Video Generator for Text, Image & Reference Videos
SP010 Hailuo AI Hailuo AI Video & Image Generator for Creators Online
SP011 Wan AI Wan AI: Leading AI Video Generation Model
SP012 Jimeng 即梦AI - 即刻造梦
SP013 Pika Pika Give any existing agent the ability to make rich content with access to the best creative models and Pika-made skills.
SP014 Artificial Analysis Video Model Comparisons
SP015 Artificial Analysis Text to Video Leaderboard - Top AI Video Models Dreamina Seedance 2.0 720p currently leads among Text to Video models with audio output in the Artificial Analysis Text to Video Arena with an Elo score of 1222.
SP016 Artificial Analysis Image to Video Leaderboard - Top AI Video Models
SP017 Andreessen Horowitz The Top 100 Gen AI Consumer Apps – 5th Edition
SP018 TechCrunch AI video startup Runway raises $315M at $5.3B valuation, eyes more capable world models
SP019 TechCrunch China’s Moonshot AI raises $2B at $20B valuation as demand for open source AI skyrockets
SP020 China Law Translate Interim Measures for the Management of Generative Artificial Intelligence Services
SP021 State Council Information Office / Xinhua China requires labeling of AI-generated online content
SP022 Loeb & Loeb LLP China’s AI-Labeling Measures and Mandatory National Standards Take Effect September 1
SP023 Harris Sliwoski LLP China’s New AI Labeling Rules: What Every China Business Needs to Know
SP024 National Institute of Standards and Technology AI Risk Management Framework
SP025 Bureau of Industry and Security Homepage | Bureau of Industry and Security
SP026 Greenberg Traurig LLP Enforcement Pause Has Limits: BIS Clarifies Ongoing License Requirement for Advanced Computing Items to China-Linked Entities
SP027 European Commission AI Act
SP028 Future of Life Institute Article 53: Obligations for Providers of General-Purpose AI Models
SI001 AIsphere 爱诗科技AIsphere 爱诗科技致力于打造全球领先的AI视频生成大模型及应用。
SI002 PixVerse PixVerse Platform - One of the best AI video API provieded by PixVerse PixVerse Platform - One of the best AI video API provieded by PixVerse
SI003 PixVerse Platform Docs Pricing - PixVerse Platform Docs $1 = 5 videos (v6, 720p, 5s, no audio, with Starter pack)
SI004 PixVerse Blog PixVerse Evolves From Creation Tool to Full Production Platform One subscription covers the team. Credits are pooled and tracked at the organizational level, with usage visibility per team member.
SI005 PixVerse Blog PixVerse R1: 720p Real-Time Video API and Partner Program Now Open This isn’t a general availability launch — it’s a selective partnership model designed to give qualified teams early access, favorable pricing, and direct influence on the R1 API’s roadmap.
SI006 PixVerse Blog PixVerse Canvas: The Visual Workspace Behind Every AI Video Project Cost is measured in credits, and the exact number depends on the model, inputs, and plan.
SI007 PixVerse Blog PixVerse Updates R1: Personalized Avatars and Shared Worlds Arrive These changes represent R1’s evolution from a real-time video generation engine into a platform for persistent, collaborative digital experiences.
SI008 PixVerse PixVerse | Create Amazing AI Videos from Text & Photos with AI Video Generator PixVerse R1 API Platform Earn Credits PixVerse AI Video Generator
SI009 PixVerse PixVerse App Download: Get the AI Video Generator App for Mobile Text-to-Video & Image Animation Start your PixVerse App Download and create AI videos on mobile with text-to-video, image animation, cinematic effects, creative templates, social clips and fast video generation for creators anywhere.
SI010 Google Play PixVerse: AI Video Generator - Apps on Google Play Join millions of creators worldwide and redefine storytelling with PixVerse.
SI011 Andreessen Horowitz The Top 100 Gen AI Consumer Apps – 5th Edition From the last mobile Brink List, two companies—PolyBuzz and Pixverse—moved into the core rankings.
SI012 CnTechPost AI video startup AIsphere raises $300 million in record China funding The startup’s annual recurring revenue (ARR) for 2025 exceeded $40 million.
SI013 Yicai Global via r.jina.ai AIsphere Raises USD300 Million, Most by a Chinese Text-to-Video Startup, Report Says The company has more than USD40 million in annual recurring revenue at the end of last year, and the user base of its app, called PixVerse overseas and Paiwo AI in China, exceeded 100 million in October, with more than 16 million monthly active users.
SI014 The AI Insider AIsphere Raises $300M to Scale PixVerse AI Video Platform and Global Expansion AIsphere reported that PixVerse has surpassed 100 million users across 175 countries, with more than 16 million monthly active users and over $40 million in annual recurring revenue.
SI015 Longbridge / Yicai AIsphere Raises USD300 Million, Most by a Chinese Text-to-Video Startup, Report Says AIsphere’s app, PixVerse, has over 100 million users, and the company is considering evolving its technology into an AI-native video game engine.
SI016 KrASIA AIsphere touts PixVerse as the “Canva for video generation,” and lands the funding to prove it Founder and CEO Wang Changhu told 36Kr that subscription revenues from its products already cover costs.
SI017 PR Newswire PixVerse Raised $60M Series B to Accelerate Global AI Video Adoption PixVerse now serves more than 100 million users worldwide, who have collectively produced over 800 million videos.
SI018 Securities and Exchange Commission cloud-20251231 As of December 31, 2025, the aggregate amount of the transaction price allocated to remaining performance obligations was $2,495.8 million.
SI019 Securities and Exchange Commission ddog-20251231 Our free cash flow was $914.7 million, $775.1 million and $597.5 million for the years ended December 31, 2025, 2024 and 2023, respectively.
SI020 Securities and Exchange Commission snow-20260131 Our net revenue retention rate, which was 125% as of January 31, 2026.
SI021 Securities and Exchange Commission nvda-20260125 Building on the architectural breakthroughs of Blackwell and leveraging Dynamo inference software, it delivers a significant increase in token throughput and reduction in cost per token compared to the Hopper generation.
SI022 Securities and Exchange Commission adbe-20251128 Cost of revenue of $2.55 billion during fiscal 2025 increased by $193 million, or 8%, compared to fiscal 2024.
SI023 Adobe ADBE 10K FY25 - FOR PRINTING ONLY, DO NOT USE Cash flows from operations of $10.03 billion during fiscal 2025 increased by $1.98 billion, or 25%, compared to fiscal 2024.
SI024 Runway AI Image and Video Pricing from $12/month | Runway AI Enterprise For teams scaling AI video production. Contact Sales.
SI025 TechCrunch AI video startup Runway raises $315M at $5.3B valuation, eyes more capable world models AI video-generation startup Runway has raised a $315 million Series E round, nearly doubling its valuation to $5.3 billion.
SI026 Artificial Analysis Video Model Comparisons Compare quality Elo, speed, and pricing across text to video, image to video, and audio-enabled video models.
SE001 PixVerse Frontier AI Research and Products that Redefine the Future of Video Intelligence
SE002 PixVerse PixVerse | Create Amazing AI Videos from Text & Photos with AI Video Generator
SE003 PixVerse Blog Download PixVerse App For Free: AI Video Generator on iOS & Android
SE004 PixVerse Platform Docs Pricing - PixVerse Platform Docs
SE005 PixVerse Platform PixVerse Platform - One of the best AI video API provieded by PixVerse
SE006 PixVerse Blog PixVerse Launches V6: Advancing AI Video Generation With Cinematic Precision
SE007 PixVerse Blog PixVerse Launches R1: The First Real-Time AI World Model
SE008 PixVerse Blog PixVerse Evolves From Creation Tool to Full Production Platform
SE009 PixVerse Blog PixVerse Canvas: The Visual Workspace Behind Every AI Video Project
SE010 PixVerse Blog PixVerse Updates R1: Personalized Avatars and Shared Worlds Arrive
SE011 PixVerse Blog PixVerse R1: 720p Real-Time Video API and Partner Program Now Open
SE012 PixVerse Blog PixVerse V6 Review: AI Video Generation Reaches Production Grade
SE013 PixVerse Blog PixVerse Meets Tripo Studio for Faster Game Asset Look-Dev
SE014 PixVerse Blog PixVerse Joins KAGAMI Gate IP Licensing PoC With Captain Tsubasa
SE015 PixVerse Blog Captain Tsubasa Joins PixVerse: Turn Your Photos Into Football Highlight Videos
SE016 PixVerse Blog PixVerse Partners with UN AI for Good Global Summit 2026 and Film Festival
SE017 PixVerse PixVerse V5.6 Ranks #2 on Artificial Analysis Leaderboard
SE018 Artificial Analysis Image to Video Leaderboard - Top AI Video Models
SE019 Artificial Analysis Text to Video Leaderboard - Top AI Video Models
SE020 Google Play PixVerse: AI Video Generator - Apps on Google Play
SE021 AppBrain PixVerse: AI Video Generator APK Download
SE022 Product Hunt (archived by Internet Archive) PixVerse: Use AI to generate fantastic character and landscape videos.
SE023 China Law Translate Interim Measures for the Management of Generative Artificial Intelligence Services
SE024 Loeb & Loeb LLP China’s AI-Labeling Measures and Mandatory National Standards Take Effect September 1
SE025 Xinhua / State Council Information Office China requires labeling of AI-generated online content
SE026 Bureau of Industry and Security Homepage | Bureau of Industry and Security
SE027 Greenberg Traurig LLP Enforcement Pause Has Limits: BIS Clarifies Ongoing License Requirement for Advanced Computing Items to China-Linked Entities
SE028 Harris Sliwoski LLP China’s New AI Labeling Rules: What Every China Business Needs to Know
SE029 OpenAI Sora is here
SE030 Apptopia PixVerse: AI Video Generator - app store revenue, download estimates, usage estimates and SDK data
SU001 PixVerse PixVerse | Create Amazing AI Videos from Text & Photos with AI Video Generator
SU002 PixVerse PixVerse App Download: Get the AI Video Generator App for Mobile Text-to-Video & Image Animation
SU003 Google Play PixVerse: AI Video Generator - Apps on Google Play
SU004 AppBrain PixVerse: AI Video Generator APK Download
SU005 Apptopia PixVerse: AI Video Generator - app store revenue, download estimates, usage estimates and SDK data | Apptopia
SU006 Uptodown PixVerse (Android)
SU007 APKPure PixVerse APK for Android Download
SU008 Product Hunt via Wayback PixVerse: Use AI to generate fantastic character and landscape videos. | Product Hunt
SU009 PixVerse Platform PixVerse Platform - One of the best AI video API provieded by PixVerse
SU010 PixVerse Platform Docs Pricing - PixVerse Platform Docs
SU011 PixVerse Blog PixVerse Evolves From Creation Tool to Full Production Platform
SU012 PixVerse Blog PixVerse R1: 720p Real-Time Video API and Partner Program Now Open
SU013 PixVerse Blog PixVerse Canvas: The Visual Workspace Behind Every AI Video Project
SU014 PixVerse Blog How to Get Free PixVerse R1 Invite Codes in 2026
SU015 PixVerse Blog PixVerse R1 Resolution Guide: How to Get 720p and 1080p Quality
SU016 PixVerse Blog Viral AI Dancing Baby: Create Trending AI Dance Videos with PixVerse
SU017 PixVerse Blog Winter Sovereign: Rule the Frozen Kingdom with PixVerse AI Video Template
SU018 PixVerse Blog Experience Fly to the Sun Moment with PixVerse AI Video Template
SU019 PixVerse Blog Captain Tsubasa Joins PixVerse: Turn Your Photos Into Football Highlight Videos
SU020 PixVerse Blog PixVerse Joins KAGAMI Gate IP Licensing PoC With Captain Tsubasa
SU021 PixVerse Blog PixVerse Meets Tripo Studio for Faster Game Asset Look-Dev
SU022 PixVerse Blog PixVerse Partners with UN AI for Good Global Summit 2026 and Film Festival
SU023 PR Newswire PixVerse Raised $60M Series B to Accelerate Global AI Video Adoption
SU024 CnTechPost AI video startup AIsphere raises $300 million in record China funding
SU025 Yicai Global via r.jina.ai AIsphere Raises USD300 Million, Most by a Chinese Text-to-Video Startup, Report Says
SU026 The AI Insider AIsphere Raises $300M to Scale PixVerse AI Video Platform and Global Expansion
SU027 Andreessen Horowitz The Top 100 Gen AI Consumer Apps – 5th Edition
SU028 KrASIA AIsphere touts PixVerse as the “Canva for video generation,” and lands the funding to prove it
SU029 PixVerse Blog PixVerse V6 Review: AI Video Generation Reaches Production Grade
SR001 AIsphere 爱诗科技AIsphere
SR002 PixVerse Platform PixVerse Platform - One of the best AI video API provieded by PixVerse
SR003 PixVerse Platform Docs Pricing - PixVerse Platform Docs
SR004 PixVerse Blog PixVerse Launches R1: The First Real-Time AI World Model
SR005 PixVerse Blog PixVerse Evolves From Creation Tool to Full Production Platform
SR006 PixVerse Blog PixVerse R1: 720p Real-Time Video API and Partner Program Now Open
SR007 Artificial Analysis Video Models
SR008 PixVerse 重塑未来视觉内容的 前沿 AI 研究与产品 | PixVerse
SR009 AppBrain PixVerse: AI Video Generator APK Download
SR010 APKPure PixVerse APK for Android Download
SR011 OpenAI Help Center What to know about the Sora discontinuation
SR012 Runway Runway | Building AI to Simulate the World
SR013 CnTechPost AI video startup AIsphere raises $300 million in record China funding
SR014 Sina Finance 爱诗科技与阿里云达成全栈AI合作 阿里云将为爱诗科技提供涵盖基础设施及大模型服务在内的全栈AI支持。
SR015 China Law Translate Interim Measures for the Management of Generative Artificial Intelligence Services These measures apply to the use of generative AI technologies to provide services to the public in the PRC for the generation of text, images, audio, video, or other content.
SR016 Regulations.AI Interim Measures for the Administration of Generative AI Services
SR017 State Council Information Office China requires labeling of AI-generated online content The guidelines are set to take effect on Sept. 1.
SR018 Loeb & Loeb LLP China’s AI-Labeling Measures and Mandatory National Standards Take Effect September 1
SR019 Harris Sliwoski LLP China’s New AI Labeling Rules: What Every China Business Needs to Know
SR020 Greenberg Traurig LLP BIS Clarifies Ongoing License Requirement for Advanced Computing Items to China-Linked Entities A BIS export license is required to export advanced computing items to entities headquartered in Country Group D:5.
SR021 Bureau of Industry and Security Homepage | Bureau of Industry and Security
SR022 EU AI Act Article 53: Obligations for Providers of General-Purpose AI Models
SR023 EU AI Act High-level summary of the AI Act
SR024 European Commission AI Act
SR025 NIST AI Risk Management Framework
SR026 Norton Rose Fulbright AI in litigation series: An update on AI copyright cases in 2026
SR027 Cleary IP & Tech Insights The Open Questions in U.S. Generative AI Copyright Litigation
SR028 AIMultiple Generative AI Copyright: Law, Litigation & Best Practices in 2026
SR029 U.S. Securities and Exchange Commission NVIDIA Corporation Form 10-K (FY ended January 25, 2026)
SR030 U.S. Securities and Exchange Commission Cloudflare, Inc. Form 10-K (FY ended December 31, 2025)
SR031 U.S. Securities and Exchange Commission Datadog, Inc. Form 10-K (FY ended December 31, 2025)
SR032 U.S. Securities and Exchange Commission Snowflake Inc. Form 10-K (FY ended January 31, 2026)
SR033 PixVerse Blog PixVerse Joins KAGAMI Gate IP Licensing PoC With Captain Tsubasa
SR034 PixVerse Blog Captain Tsubasa Joins PixVerse: Turn Your Photos Into Football Highlight Videos
SR035 PixVerse Blog PixVerse Meets Tripo Studio for Faster Game Asset Look-Dev
SR036 LexisNexis China Interim Measures for the Management of Generative Artificial Intelligence Services
SR037 Regulations.AI Measures for the Identification of AI-Generated (Synthetic) Content (人工智能生成合成内容标识办法)
SR038 Comparative AI Interim Measures for the Management of Generative AI Services
SV001 CnTechPost AI video startup AIsphere raises $300 million in record China funding
SV002 Yicai Global via r.jina.ai AIsphere Raises USD300 Million, Most by a Chinese Text-to-Video Startup, Report Says
SV003 The AI Insider AIsphere Raises $300M to Scale PixVerse AI Video Platform and Global Expansion
SV004 KrASIA AIsphere touts PixVerse as the “Canva for video generation,” and lands the funding to prove it
SV005 PR Newswire PixVerse Raised $60M Series B to Accelerate Global AI Video Adoption
SV006 AIsphere 爱诗科技AIsphere
SV007 PixVerse Platform Docs Pricing - PixVerse Platform Docs
SV008 Artificial Analysis Video Model Comparisons
SV009 Artificial Analysis Text to Video Leaderboard - Top AI Video Models
SV010 Artificial Analysis Image to Video Leaderboard - Top AI Video Models
SV011 Runway AI Image and Video Pricing from $12/month | Runway AI
SV012 TechCrunch AI video startup Runway raises $315M at $5.3B valuation, eyes more capable world models | TechCrunch
SV013 Sacra Runway revenue, valuation & funding
SV014 StartupHub.ai Runway Valued at $5.3 Billion
SV015 Pika Pika
SV016 Sacra Pika valuation, funding & news
SV017 FirmKnow Pika's Valuation Jumps to 3.4B After Securing 80M in Funding for AI Video Tech - FirmKnow
SV018 Startup Wired Zhipu AI Targets $640M Hong Kong IPO, $6.7B Valuation
SV019 Business Day / Reuters Chinese AI firm MiniMax joins year-end listings rush in Hong Kong
SV020 WinBuzzer MiniMax Raises $619M in Hong Kong IPO as Chinese AI Startups Beat Silicon Valley to Public Markets
SV021 The AI Rankings Moonshot AI in 2026: Kimi, K2.6, $20B Valuation & Strategy | The AI Rankings
SV022 TechCrunch China's Moonshot AI raises $2B at $20B valuation as demand for open source AI skyrockets | TechCrunch
SV023 Business 2.0 News Moonshot AI $20B Valuation 2026: China's Largest LLM Raise Reshapes AI Race
SV024 Securities and Exchange Commission adbe-20251128
SV025 Securities and Exchange Commission nvda-20260125
SV026 Securities and Exchange Commission snow-20260131
SV027 Securities and Exchange Commission ddog-20251231
SV028 Securities and Exchange Commission cloud-20251231
SV029 Longbridge / Yicai AIsphere Raises USD300 Million, Most by a Chinese Text-to-Video Startup, Report Says
SV030 Runway Runway | Building AI to Simulate the World