← Video

ViduVidu

生数科技（清华系）Shengshu Technology (Tsinghua-linked)

免费+订阅Free + subscription 国产直连Direct access in China 多主体参考Multi-subject references 动画向Animation-focused 清华系Tsinghua origins partial lifecycle: active 核验Checked 2026-10-22

Vidu 是清华系生数科技的视频大模型入口：招牌是「参考生视频」多主体一致性—— 角色一致性要求高的叙事/动画向场景，这是它公开说明的产品方向。

Vidu is the Tsinghua-linked Shengshu Technology's video model entry: its signature is reference-to-video with multi-subject consistency — for narrative/animation work that needs character consistency, it's the one domestic player leaning hardest on that.

适合与不适合When to use it, when not to

适合：多角色叙事一致性、动漫风格动态化、国产直连使用。不适合：模型自选、开放权重、写实大场面主打。

Fits: multi-character narrative consistency, anime-style motion, domestic direct access. Not a fit: model self-selection, open weights, photoreal big-scene work.

中文生态China availability 国产 / 国内可直接用 / 备案状态 · 快照domestic / direct access / CAC filing · snapshot 2026-10-22

出品属地Origin: 国产domestic　北京生数科技有限公司（中国，清华系团队）。Beijing Shengshu Technology Co., Ltd. (China, Tsinghua-linked team).
国内使用Direct access: 大陆可直连direct OK　vidu.cn / vidu.studio 大陆直连实测 200（2026-10-22）。vidu.cn / vidu.studio measured 200 direct from mainland (2026-10-22).
备案状态CAC filing: 已备案Filed：Beijing-Vidu-20241125S0003Beijing-Vidu-20241125S0003　·　网信办 2025-01 批次公示直抓核验（2026-10-22）；另有算法备案「Vidu视频生成算法-1/-2」（阿里云百炼公示页交叉核验）。verified against the CAC January 2025 published batch (direct fetch, 2026-10-22); the algorithm filings "Vidu video generation algorithm -1/-2" are cross-verified via Alibaba Cloud Bailian's disclosure page.

备案信息以网信办《生成式人工智能服务已备案信息》公示名单为准（按批次滚动更新，本站只抄编号不解读）；「大陆可直连」是本站从大陆网络直连实测的结果，不是厂商承诺。GitHub 类站点当日实测有波动，口径写在依据里。Filing status is quoted from the CAC public registry (updated batch by batch; this site copies the number and does not interpret it). "Direct access" is this site's own reachability measurement from a mainland network, not a vendor promise. GitHub-hosted sites fluctuated on the measurement day — the note says so.

固定坐标系Fixed coordinate system

8 个维度dimensions, 与同分区其他对象逐项可比each comparable item by item against others in the same section.

模型与开放条件Model & openness: 自研 Vidu 视频大模型线（官方 FAQ：北京生数科技联合清华大学发布，主打长时长/高一致性/高动态）；用户不选裸模型。开放权重线未核验到官方声明。
In-house Vidu video large-model line (official FAQ: released by Beijing Shengshu Technology jointly with Tsinghua University, leading on long duration/high consistency/high dynamics); users do not pick bare models. An open-weights line — no official statement verified.
运行位置Where it runs: 网页（vidu.cn 国内入口 / vidu.com 海外入口）+ API（企业侧，官方叙事）；生成在厂商云端。
Web (vidu.cn domestic entry / vidu.com overseas entry) + API (the enterprise side, official narrative); generation on the vendor's cloud.
本地文件Local files: 作品存账户侧云端，生成历史在网页作品页可见；下载归本地。未见官方声明的本机数据面。
Works stored account-side in the cloud, generation history visible on the web works page; downloads go local. No officially stated local data surface.
关机后的任务Tasks after shutdown: 生成即时/排队（按档位波动，未逐条核验）；无「关机续跑」任务面（未核验到声明）。
Generation is instant/queued (varies by tier, not verified item by item); no "keep-running-after-shutdown" task surface (no claim verified).
工具与扩展Tools & extensions: 文生视频、图生视频、参考生视频（多主体参考）是官方点名的差异功能（角色一致性向）；首尾帧控制线（官方 FAQ 所见）；2D 动画风格线（官方 FAQ 叙事）。
Text-to-video, image-to-video, reference-to-video (multi-subject reference) is the differentiating function named by the official side (character-consistency oriented); a first/last-frame control line (seen in the official FAQ); a 2D-animation style line (official FAQ narrative).
上下文与记忆Context & memory: 以账户作品组织；跨会话记忆未核验到官方声明。多主体参考是它的「一致性」机制——参考图复用不是记忆。
Organized by account works; cross-session memory — no official statement verified. Multi-subject reference is its "consistency" mechanism — reusable reference images are not memory.
权限与限制Permissions & limits: 账号+订阅体系（D9：价格不采）；训练用途开关未逐条核验。备案：Vidu 备案号 Beijing-Vidu-20241125S0003（网信办 2025-01 批次公示直抓核验，2026-10-22）；算法备案「Vidu视频生成算法-1/-2」见深度合成算法备案清单（阿里云百炼公示页交叉）。
Account + subscription system (D9: prices not collected); the training-use switch not verified line by line. Filing: the Vidu filing no. Beijing-Vidu-20241125S0003 (CAC 2025-01 published batch, direct-fetched and verified 2026-10-22); the algorithm filings "Vidu video generation algorithm -1/-2" (official Chinese registered names kept in the Chinese entry) appear in the deep-synthesis algorithm filing list (cross-checked on Alibaba Cloud Model Studio's disclosure page).
适合什么任务Best for: 适合：角色一致性要求高的叙事/动画向生成（多主体参考）、国产直连、动漫风格动态化。不适合：要模型自选、要开放权重、写实电影级大场面主打。
Fits: narrative/animation-oriented generation with high character-consistency demands (multi-subject reference), domestic direct connection, anime-style motion. Does not fit: model self-selection, open weights, a photorealistic cinematic set-piece focus.

头号误解Top misconception

「参考生视频」是它的招牌差异——按普通文生视频工具理解会低估多主体参考的价值'Reference-to-video' is the signature differentiator — judging it as a plain text-to-video tool misses the multi-subject consistency value
vidu.cn（国内）与 vidu.com（海外）是双入口——账号/内容政策是否互通未核验，别按同一体系理解vidu.cn (domestic) and vidu.com (overseas) are two entries — account/content policy interop not verified
备案号主语是「Vidu」服务（网信办公示）——算法备案（网信算备…号）与服务备案是两套编号，别混引The filing number's subject is the Vidu service (CAC registry) — algorithm filings (the CAC algorithm-filing numbers) and service filings are two numbering systems; don't cross-quote them

未知项清单Known unknowns

开放权重线（未核验到官方声明）
API 企业侧的接入面与配额（未逐条核验）
训练用途开关（未逐条核验）
跨会话记忆机制（未核验到官方声明）

The open-weights line (no official statement verified)
The API enterprise side's access surface and quotas (not verified item by item)
The training-use switch (not verified line by line)
Cross-session memory mechanism (no official statement verified)

证据来源Evidence sources

判断可回到以下一手源复核。本站核验日Every judgement can be rechecked against the primary sources below. Verified on 2026-10-22，内容更新日content updated on 2026-10-22。

类型Type	名称Name	链接Link
official	Vidu · 官网（产品与能力叙事、官方 FAQ）Vidu · official site (product and capability narrative, official FAQ)	https://www.vidu.cn
official	Vidu · Terms（官方条款，2026-10-22 直抓核验）Vidu · Terms (official terms, direct-fetched and verified 2026-10-22)	https://www.vidu.com/terms
changelog	Vidu · Blog（官方更新线，changelog）Vidu · Blog (official update line, changelog)	https://www.vidu.com/blog
docs	网信办 · 生成式人工智能服务已备案信息（2025-01 批次，备案号出处）CAC · filed generative-AI service information (2025-01 batch, filing-no. source)	https://www.cac.gov.cn/cms/pub/interact/downloadfile.jsp?filepath=NUtqEIwGiCjGm2Bhl20cvMJN~~~VfG4e3TRGnFD8PfRvz0vHSSJ6sATU6DtmcGo1/SlbWBI5~oROMVCft2OZdVAfqphfFn00jUfFRLdBE5I=&fText=%E7%94%9F%E6%88%90%E5%BC%8F%E4%BA%BA%E5%B7%A5%E6%99%BA%E8%83%BD%E6%9C%8D%E5%8A%A1%E5%B7%B2%E5%A4%87%E6%A1%88%E4%BF%A1%E6%81%AF(2025%E5%B9%B41%E6%9C%88)

实测记录Measured runs

本档未跑生成质量与速度测试；结论只覆盖能力面与许可面。

实测建议（三条，都指向官方可读页没给的东西）：

多主体参考在 3+ 角色场景的一致性衰减
首尾帧控制的实际插值质量
国内/海外双入口的作品互通性

This site has not completed measured runs. The coding-oriented test protocol is at protocols/coding/_protocol.md (the coding task suite does not apply to creative products).

Suggested measurements (three, all pointing at things the official readable pages do not give):

Consistency falloff of multi-subject reference in 3+ character scenes
The actual interpolation quality of first/last-frame control
Work interoperability between the domestic/overseas dual entries