Vidu

Quick Verdict

Vidu is an AI video platform from Beijing Shengshu Technology. Its official product currently centers on text-to-video, image-to-video, and reference-to-video. The distinctive workflow is not simply turning a sentence into a clip: creators can save people, objects, and scenes as reusable subjects, then use multiple reference images to constrain their appearance across shots. Vidu says reference-to-video accepts up to seven subject images. Image-to-video also supports first and last frames, asking the system to generate the motion between two endpoints.

That combination is useful for recurring characters, product demonstrations, animation, advertising shots, and storyboard previews. It does not guarantee perfectly locked identity, clothing, props, or anatomy in every frame. Production use still requires generation sampling, editing, and inspection. Buyers must evaluate four things together: whether they own the reference material, how many credits the selected mode consumes, whether a paid account grants the commercial permission needed for the project, and which visible or invisible AI labels must remain. A headline subscription price alone does not reveal the true usable-clip cost.

Best For

Vidu is best for short-form creators, animation teams, ad studios, storyboard artists, and social teams that repeatedly use the same character or product. Reference generation is less important for a single abstract atmosphere shot. Projects requiring long-form continuity, precise acting, extensive dialogue synchronization, or frame-level control should treat Vidu as a shot source, not a complete production replacement. Editing, voice, sound design, color, rights review, and quality control remain separate responsibilities.

Key Features

• Reference-to-video: uses multiple views of a person, object, or scene as subject constraints. The official site states that users can upload up to seven subject images and reuse approved assets from a subject library.
• Image-to-video and endpoints: animates a still or accepts first and last frames, allowing the model to create an intermediate transition.
• Text-to-video: describes subject, action, camera, lighting, style, and pacing without requiring an initial image.
• Animation and dynamic motion: Vidu highlights fluid animation alongside advertising, entertainment, tourism, and film examples.
• Templates and API: the consumer product includes social-effect templates, while businesses can evaluate the separate API platform. Web subscription credits should not be assumed to transfer to API billing.

Use Cases

Define the shot’s purpose, aspect ratio, duration, and features that must not change. Character references should include useful angles, clothing, and key props; scene references should not contradict one another. Save approved assets in the subject library. Use a lower-cost or off-peak mode for composition and movement tests before paying for a final specification. In later passes, keep the successful prompt fixed and change only one variable, such as camera movement or expression, so failures remain diagnosable.

For endpoint generation, check that the two frames have compatible subject counts, viewpoint, lighting, and geometry. Generate several candidates, then inspect flicker, anatomy, text, logos, subject identity, and shot continuity in an editor. Add sound and captions only after picture lock. For a recurring campaign, test ten representative shots before committing a monthly budget; one attractive homepage sample says little about consistency under your actual references.

Pricing

Vidu’s pricing page loads current plan numbers dynamically behind the application state, so this review does not repeat the old 80/240/600-credit table. The official pricing FAQ does clearly describe three credit classes. Subscription credits last 30 days. Extra purchased credits last two years and are available only to paid users. Reward credits from campaigns, competitions, or creator programs also last two years. The documented spending order uses subscription credits before purchased credits.

The site promotes free generation in an off-peak mode, but available models, queue time, duration, and output specification can vary in the account. Before paying, record the exact credit cost of the intended mode, whether failed jobs are refunded, how unused monthly credits expire, and whether purchased credits remain usable after cancellation. API concurrency and pricing require a separate check. This matters because two plans with similar headline prices can produce very different numbers of usable shots.

Vidu’s terms, updated July 3, 2026, state that Vidu does not claim ownership of User Content and users retain their rights. Users grant a worldwide, non-exclusive, royalty-free license needed to host, process, display, support, and improve the service. The terms also state that User Content is not used to train or improve AI or machine-learning models. Most importantly for buyers, the ordinary service license is personal and non-commercial, while the license extends to commercial use for paid users. Free-credit outputs therefore should not be placed directly into commercial advertising under the current wording.

Payment does not clear third-party rights. Users must have permission for uploaded people, music, characters, product images, trademarks, and client assets, and generated outputs may resemble other users’ work. Vidu does not promise uniqueness or non-infringement. Its terms also address generated-content labeling: outputs may carry visible or invisible AI identification, and users may not remove or tamper with required marks. Even where a user applies for no visible mark, invisible identification remains and the publisher may still need to disclose AI generation. Commercial teams should retain source licenses, consent records, paid-account receipts, generation dates, and a record of human edits.

Pros

Reusable subjects and multi-image references fit episodic content; first-and-last-frame input helps with planned transitions; Chinese-language consumer tools, templates, and an API cover experimentation through batch production; and official documentation explains credit expiration and commercial-license differences.

Cons

Plan numbers and per-mode credit costs are dynamic; consistency remains probabilistic under occlusion, multiple people, and small details; 30-day subscription credits can be wasted by low-frequency users; free users do not receive the paid commercial-use extension; and removing a visible watermark never removes rights or disclosure obligations.

Alternatives

FAQ

Is Vidu free?

It offers trial, login, campaign, or off-peak generation opportunities, but models and output specifications vary. Under the current terms, the free service license is personal and non-commercial.

How does reference-to-video differ from image-to-video?

Image-to-video animates a still or connects endpoint frames. Reference-to-video uses multiple images of people, objects, or scenes to constrain reusable subject identity.

Do subscription credits expire?

Yes. The official FAQ says subscription credits last 30 days. Purchased and reward credits generally last two years, subject to the current purchase rules.

Can paid outputs be used commercially?

The current terms extend the service license to commercial use for paid users, but users remain responsible for every input right, likeness, music track, brand element, and final output review.

Does no visible mark mean no AI label?

No. The terms say invisible identification remains where required, even if an application to omit the visible mark is accepted, and publishers may still have disclosure duties.

Bottom Line

Vidu should be judged by whether its subject library and references reduce rework across a sequence, not by one impressive clip. Run ten realistic shots with the same character, calculate usable output per credit, queue time, and repair time, then compare the result with Kling AI or Runway. Its reference controls are valuable only when they outweigh expiring credits, selection work, and rights administration.