Video
Vidu Q3 Mix
Video
Available on Ofrai — chat, images, and video in one place with shared credits.
Example prompts
Choose an example to load its prompt and recommended settings into the Vidu Q3 Mix creator above.
How it works
Create with Vidu Q3 Mix in three steps
- 01
Choose your inputs
Keep Vidu Q3 Mix selected, then add a prompt and any supported reference assets.
- 02
Set the output
Choose the aspect ratio, resolution, and available Video options before generating.
- 03
Generate and refine
Review the result, adjust the prompt or settings, and generate another version from the same place.
Vidu Q3 is documented as a family of three model ids behind one endpoint — viduq3-pro, viduq3-turbo and viduq3 — and the choice between them is not a quality dial you turn at the end. It changes what inputs the model accepts, how short a clip you may ask for, and what you pay. The pro variant is the high-quality tier and is documented as supporting audio-video synchronisation and storyboard video generation, at roughly twice the turbo price. Turbo is the fast, value tier with smart scene switching.
The plain viduq3 id is the one documented for multi-camera consistency, smart shot switching and multi-reference generation, and it is also the only one that treats reference images as required. All three top out at 1080p and all three generate audio by default.
Three Q3 model ids that are not interchangeable
The variant you pick decides the shape of the request, so it belongs in the brief rather than in a late-stage setting. viduq3-pro is the quality tier: the reference documents it as supporting audio-video sync and storyboard video generation, and it costs about twice as much as turbo. viduq3-turbo is the speed tier — smart scene switching, fastest generation, and the reference explicitly calls it the best value of the three.
The plain viduq3 is the consistency tier: it is documented as the strongest of the three for multi-camera consistency, with smart shot switching and multi-reference generation. Because pro and turbo accept the same image inputs, a job that begins as turbo can be promoted to pro for the final render without rewriting the request. A job that begins as plain viduq3 cannot be demoted to turbo without losing the reference handling you chose it for. That asymmetry is the practical reason to start specific.
Audio is on by default across the whole Q3 line
Every Q3 model documents audio as defaulting to true, so a request that says nothing about sound still comes back with sound. Setting it to true explicitly asks the system to generate speech and background sound effects derived from the prompt, which means the prompt text is doing double duty: it describes the shot and it feeds the audio. This is the one endpoint family in this set where silence is the thing you have to ask for. It matters for two reasons.
First, for silent deliverables — a clip destined for a cut with its own score, or a social post that will carry licensed music — you must set the flag off rather than assume silence. Second, when a render comes back with dialogue you did not ask for, the cause is usually a prompt that described a character doing something speakable, not a bug. The pro variant's documented audio-video sync support is the reason to reach for it when the sound has to line up with the picture.
Duration floors are not the same across variants
Duration defaults to 5 seconds on every Q3 model, and the documented range then diverges. For viduq3-pro and viduq3-turbo the range runs from 1 second to 16 seconds. For the plain viduq3 the range starts at 3 seconds. In practice the difference only shows up in the very short end, and that is exactly where it decides a workflow. A one-second insert — a product flash, a match cut, a transition sliver — is available to pro and turbo and unavailable to viduq3.
If a sequence is built from stinger-length clips and then you switch the whole batch to viduq3 for consistency, every sub-3-second beat has to be re-cut. Pick the variant once, at the top of the sequence, and check your shortest shot against its floor. The upper bound is 16 seconds for all three, and that number is itself a warning: past roughly ten seconds a single generation is usually doing the work of several shots.
Reference images mean two different things in this family
The image input has genuinely different semantics depending on which id you send, and this is the single most common source of wasted renders. For viduq3-pro and viduq3-turbo, image_urls is optional and the count decides the mode: send nothing for text-to-video, send exactly one and it becomes the first frame for image-to-video, send two and the first becomes the opening frame and the second the closing frame, giving you a first-and-last-frame generation.
For plain viduq3, image_urls is required and accepts up to 7 reference images — not frames. Those seven are consistency references, which is how the variant delivers its documented multi-camera consistency and smart shot switching. Sending two images to viduq3 does not give you a first-and-last-frame clip; it gives you two references and leaves the shot structure to the model. Read which contract applies before you assemble the payload.
Aspect ratio is flexible, and reference images can override it
The documented aspect ratio values are 16:9, 1:1 and 9:16, but the reference adds a note that the Q3 series supports arbitrary aspect ratios, so the three listed values are the common ones rather than the exhaustive set. More importantly, the request also accepts the literal value image_urls, and when you pass that the ratio is normally determined by the reference image instead of by you.
That is a useful behaviour for matching an existing asset: hand it a 4:5 still and it composes for 4:5 rather than letterboxing. It is also a trap when you care about the final frame. If the deliverable is vertical video for a specific placement, set 9:16 explicitly rather than relying on inference from a source image whose proportions came from somewhere else in the workflow. The inference follows the reference you sent; it does not follow the placement you intend.
The 5000-character prompt and @name subject references
The prompt is the only required field on this endpoint, and it accepts up to 5000 characters — a generous ceiling that invites a different writing style from short-prompt models. The documented example is deliberately shaped: a subject doing something, a camera move, then a texture word, in the form of a cat playing piano, the camera pushing in slowly, cinematic quality. Follow that order.
Subject and action first, because that is what the model resolves first; camera second, because it applies to the whole clip; style last, because it modulates rather than specifies. The Q3 series also supports named subjects, which you address inside the prompt with the @name syntax. That is how you keep one character consistent across a multi-shot request without repeating a full physical description each time.
Note that the reference does not document how subjects are declared, so treat the @name form as the referencing half of the feature rather than a complete tutorial.
FAQ
- Does Vidu Q3 generate audio if I do not ask for it?
- Yes. Audio is documented as defaulting to true across all three Q3 model ids, and setting it to true asks the system to generate speech and background sound effects from the prompt. If you need a silent clip, turn the flag off explicitly.
- How many reference images can I pass?
- It depends on the model id. For viduq3-pro and viduq3-turbo, image_urls is optional: none for text-to-video, one as the first frame, two as first and last frame. For plain viduq3, image_urls is required and accepts up to 7 reference images used as consistency references rather than frames.
- What is the shortest clip each variant will generate?
- Duration defaults to 5 seconds. viduq3-pro and viduq3-turbo accept 1 to 16 seconds; plain viduq3 accepts 3 to 16 seconds. Sub-3-second inserts are therefore only available on the pro and turbo variants.
- How much more does viduq3-pro cost than viduq3-turbo?
- The reference prices the pro quality tier at roughly twice turbo. Turbo is documented as the fastest and the best-value option, while pro is the one that adds audio-video sync and storyboard video generation.
FAQ
- Can I use Vidu Q3 Mix directly on this page?
- Yes. The creator above is the working Ofrai studio with Vidu Q3 Mix selected by default.
- Where are my finished results saved?
- Signed-in generations are kept in your Ofrai library so you can download or revisit them.
- Which prompts work best with Vidu Q3 Mix?
- Describe the subject, setting, visual style, lighting, and camera or motion clearly. The examples above provide practical starting points.
- Can I edit the example prompts before generating?
- Yes. Load any example, then change its wording, aspect ratio, references, or other supported settings in the creator.
Related models
Compare other Video models that can handle similar creative work to Vidu Q3 Mix.
Related guides
Tools
Commercial use depends on the selected model and model terms.





