What is Vidu Q4?
Vidu Q4 Preview is an AI video generation model designed for short, cinematic scenes that require consistent characters, coordinated audio and video, and controlled camera movement. Unlike a workflow that animates only one still image, its Reference-to-Video mode can combine multiple visual references and audio clips to guide the generated scene.
The model is particularly relevant to AI short dramas, character-led storytelling, product advertising, social video, and cinematic previsualization.
Main features of Vidu Q4
Reference-guided subject consistency
The reference-to-video workflow accepts 1–15 images. These can provide visual guidance for characters, products, locations, costumes, or scene styles. This is useful when a video needs to retain recognizable visual details across a sequence rather than inventing every frame independently.
Consistency is a generation objective, not a guarantee that every frame will perfectly preserve faces, proportions, logos, or small details.
Native audio and synchronized video
Q4 Preview supports audio-video output, allowing creators to generate a clip with sound instead of always adding audio in a separate editing step.
The reference-to-video endpoint also accepts up to three audio references. These can help guide audio-related aspects of a scene, depending on the supplied materials and prompt. Reference audio should not be confused with a guarantee of exact speech transcription, voice identity, or perfect lip synchronization.
Intelligent camera switching
The model supports automatic camera switching, which can help create scenes with changing viewpoints rather than a single static camera angle.
This is especially relevant to dialogue scenes, character interactions, cinematic sequences, and short promotional videos. However, creators should inspect generated cuts for continuity, subject positioning, and whether the selected angles match the intended story.
Multiple output resolutions and aspect ratios
The official model map lists output options from 540p to 4K at 24 fps. Supported aspect ratios include landscape, portrait, and square formats.
This gives creators flexibility when preparing videos for different placements, including:
- 16:9 for conventional landscape video.
- 9:16 for vertical short-form content.
- 1:1 for square social media assets.
- 3:4 and 4:3 for other portrait or landscape layouts.
Higher resolution does not automatically imply better motion, stronger consistency, or fewer visual artifacts.
Asynchronous API generation
The documented API follows a task-based workflow: submit a generation request, receive a task ID, and query the task until the generated result becomes available.
This is suitable for production pipelines that need to process multiple clips, track job status, and retrieve results programmatically.
Vidu Q4 vs. Vidu Q3 vs. other video models
| Criterion | Vidu Q4 Preview | Vidu Q3 | Other video models |
|---|---|---|---|
| Primary positioning | Reference-driven cinematic scenes | Narrative video generation | Depends on model |
| Visual references | Up to 15 images | Depends on API mode | Depends on model |
| Audio references | Up to 3 clips | Depends on model and mode | Depends on model |
| Maximum duration | 16 seconds | Up to 16 seconds on documented Q3 routes | Varies |
| Maximum resolution | 4K | Depends on route | Varies |
| Best evaluation target | Character, voice, camera and scene consistency | Narrative and audio-video workflows | Task-specific |
This is a capability comparison, not a benchmark ranking. For technical details on Vidu Q3, see CometAPI's Vidu Q3 guide
Recommended use cases
- AI short dramas: Maintain character appearance and voice across short narrative scenes.
- Advertising and product videos: Guide product appearance and scene composition with reference images.
- Social media content: Generate short cinematic clips for creative testing and campaign variations.
- Film previsualization: Explore camera movement, action sequences, and shot transitions before production.
- Character-driven animation: Combine visual references with voice guidance for more consistent performances.