Vidu AI turns text prompts, still images and up to seven reference images into short AI videos, and its Q3 model adds native audio to clips of up to 16 seconds. It is aimed at independent creators, animation teams and marketers who need characters to stay consistent across shots.
Used this tool? Rate it
Used this tool? Rate it
Vidu AI is a web-based AI video generator that turns written prompts, still images and reference images into short video clips. The product supports three main creation modes: Text to Video, Image to Video and Reference to Video. Its current flagship video model, Vidu Q3, generates sound and picture together, so a single clip can include dialogue, voiceover, sound effects and music.
The international site at vidu.com is operated by ShengShu AI HK Limited, the company named in its privacy policy and the developer shown on its Android app listing. Independent tech media describe the developer, ShengShu Technology, as a company founded in March 2023 by a team from Tsinghua University's Institute for AI Industry Research. The site and its terms do not explain how the Hong Kong entity relates to the Beijing-based ShengShu described in press coverage, so this page does not draw that link.
Three things usually decide whether Vidu fits: its focus on keeping characters and objects consistent across shots, commercial rights that only come with a paid plan, and terms that give Vidu a perpetual licence to user content, including for AI training, with subscriptions sold as non-refundable.
The number of reference images a user needs is described inconsistently on the official site. The Reference to Video page states the mode supports up to seven references and tells users to upload one to seven images, while the homepage FAQ says Reference to Video lets users upload three or more images of an object, character or scene. Both describe combining several characters, objects or scenes in one clip; they differ on the minimum.
Vidu's own guidance on choosing between the two image-driven modes is short. If you focus on storytelling and consistency, the company recommends Reference to Video; if you want to animate a single image or keyframes, it points to Image to Video.
Vidu Q3 is described by the company as a new-generation model that creates video with native audio, meant to be ready to publish without adding sound in a separate step. A single Q3 generation can run up to 16 seconds. The model page lists English, Japanese and Chinese as the supported languages for video output, and highlights multi-speaker dialogue and frame-level camera control for directing pacing.
Vidu Agent is positioned as a marketing agent built on the open-source OpenClaw framework and powered by Vidu Q3. Instead of producing one asset at a time, it is described as taking a goal such as a product video or a campaign and planning the storyboard, script and final clip. The agent runs inside the Vidu platform, so no local installation is needed.
The same generation capabilities are packaged as Vidu Skills, which the company says can also be used in OpenClaw, Claude Code or other compatible agent workflows. Setup requires installing the skills from GitHub and setting a Vidu token as an environment variable.
The official Reference to Video walkthrough follows three steps:
Every generation consumes credits, and Vidu sets the cost of each task by the type of generation, its length, its resolution and the chosen model. Saved references from My References shorten step 1 when the same character returns in later videos.
The company's own FAQ lists a mix of social, studio and commercial uses:
For narrative work, the Q3 model page names comic and manga-style drama, cinematic shots, short-form series and narrative ads as the target formats, where continuity and timing matter most. These are vendor-described uses rather than independently measured outcomes.
Two conditions narrow the fit. Subscriptions are tied to individual accounts and cannot be shared, so each person generating on a paid plan needs their own account. Commercial use is also granted only to paying users, which matters for client work and ads.
The Google Play content rating and the terms of service set different age expectations: the store lists the app as rated Everyone, while the terms require users to be at least 18 or of legal age in their jurisdiction.
Vidu pricing for the web product uses four tiers. The captured pricing page defaulted to yearly billing, and the monthly-billing prices were not displayed in that state; the figures below are the per-month rates shown with annual billing.
| Plan | Price shown (yearly billing) | Monthly credits | Stated video volume | Reference or subject quota |
|---|---|---|---|---|
| Free | $0/month | Not shown | Not shown | 10/month |
| Standard | $8/month, billed yearly as $96 | 800 | Up to 200 videos | 50/month |
| Premium | $28/month, billed yearly as $336 | 4,000 | Up to 1,000 videos | 100/month |
| Ultimate | $79/month, billed yearly as $948 | 8,000 | Off-Peak Mode: up to 200 videos submitted daily | 300/month |
The page states that yearly Standard saves $24 compared with monthly billing, Premium saves $84 and Ultimate saves $240. A banner on the page also says plans will be upgraded on October 6, so the plan details shown here may change after that date.
Each paid card lists the same core generation features: 3–16 second videos, generating up to four videos at once, a professional prompt editing mode, higher-quality H.265 Pro and H.264 Pro encoding, and a statement that generated content can be used commercially. The Ultimate plan also includes unlimited Q2 1080p image generation. The comparison table includes a "Download without watermark" row, but the plan text captured for this page does not spell out which tiers it applies to.
The API platform uses its own pay-as-you-go credit system, separate from web subscriptions, with self-serve top-ups and an enterprise channel for larger or custom needs. Its packs are sold with different validity periods and concurrency limits. For example, the Starter Pack was listed at $144 with 30,000 credits, at least five concurrent requests and six months of validity, displayed next to a $150 figure and a limited-discount label.
Independent coverage places Vidu among a group of Chinese AI video generators that includes Kuaishou's Kling AI and MiniMax's Hailuo AI, all competing with international models such as OpenAI's Sora and Google's Veo. Reporting on the Vidu Q2 release framed it as a challenge to Sora 2 and Veo 3.1.
The feature overlap is real: one report on Vidu Q2 noted that start-and-end-frame transitions are also available in Veo 3.1. Launch coverage singled out Vidu's focus on efficiency, consistency across formats and a specialization in anime-style videos as its angle against these rivals.
Vidu does not claim ownership of content you create or upload, and you keep the rights to it. That ownership comes with a wide licence in Vidu's favour: by submitting content you grant Vidu and its affiliates a worldwide, royalty-free, perpetual and irrevocable licence, which explicitly includes the right to use your content to improve and develop Vidu's products, including artificial intelligence training. The licence continues after you close your account. The privacy policy itself does not describe an opt-out from training, and none was found in the terms, privacy policy or help centre pages reviewed for this page.
Commercial rights depend on the plan. The terms grant a personal, non-commercial licence to all users and extend it to commercial use only for paid users. The help centre likewise states that only subscribed users may use Vidu-generated works commercially unless the licensing terms say otherwise.
On data handling, the privacy policy says user information is stored in Singapore. It may also be transferred across borders within the Vidu group and to third-party service providers. The listed processors include NetEase Review for content review. MiniMax is listed for generating video dubbing, voiceover or subtitles. Operational data is retained for no more than six months unless the law or a legal claim requires longer. Users may also have the right to request deletion of their information, subject to legal exceptions.
Partly. All users receive a set amount of free credits, and the Free plan includes 10 saved references per month. Commercial use of the output requires a subscription, and the monthly credit allowances in the pricing table belong to the paid plans.
Only on a paid plan. The terms limit free users to personal, non-commercial use, and each paid plan card states that generated content can be used commercially.
The terms give Vidu a perpetual, irrevocable licence to user content that includes the right to use it for artificial intelligence training. No training opt-out was found in the pages reviewed.
Vidu Q3 generates up to 16 seconds in a single run, and the paid plans list support for 3–16 second videos.
The pricing page says refunds are not provided; you can cancel before the next billing cycle to stop further charges. The one exception described is credits for a generation that failed because of a problem on Vidu's side.
Yes. According to the help centre, you keep access to existing creations after cancelling, but future generation tasks are affected by your plan type.
