Vidu Q4 API
Vidu Q4 官方旗舰预览模型，支持单图生视频与多参考生视频，提供 3～16 秒、540p 至 4K 输出、自动分镜和音画同出。
What is Vidu Q4 Preview?
Vidu Q4 Preview is Vidu's preview flagship video model. Its official API documents image-to-video and reference-to-video routes, generated audio, and automatic scene cuts. Those are vendor claims, not Roko benchmark results.
On Roko, use the exact public model ID viduq4-preview through the asynchronous /v1/videos/tasks endpoint. Image-to-video needs one image; reference-to-video needs 1–15 images and can accept up to three MP3 references.
Both routes support 3–16 seconds and 540p, 720p, 1080p, 2K, or 4K. The page remains a noindex preview until four independent examples and two Roko-run examples are available.
Model pricing
Pricing follows the current production configuration and the billing dimensions supported by each route.
| Route | Price | Billing rule |
|---|---|---|
| 基础视频 | 1080p $0.1177/sec2k $0.1864/sec4k $0.3827/sec540p $0.04415/sec720p $0.09321/sec | 按输出秒数估算；完成后按上游实际 credits 结算 |
| 生成音频 | 开启音频 $0.07359/request | 每个任务固定附加 |
Cost examples
These are fallback estimates with the 10% service fee. Successful tasks settle on returned credits × $0.004906.
5 × 19 credits × $0.004906 = $0.4661
Set generate_audio to false; the 15-credit audio add-on is excluded.
(8 × 24 + 15) credits × $0.004906 = $1.0155
The audio add-on applies once per task, not once per second.
Model information
Only provider-documented routes, capabilities, and limits implemented by this gateway are listed.
viduq4-preview
- Limits
- Images 1
- Duration
- 3–16s
viduq4-preview
- Limits
- Images 1–15 · Audio 0–3
- Duration
- 3–16s
Core capabilities
Animate one image
Turn one starting image and a prompt into a 3–16 second clip for character, product, or scene motion.
Multi-reference generation
Use 1–15 images to constrain identity, setting, and style, plus up to three MP3 sound references.
Five resolution tiers
Choose 540p, 720p, 1080p, 2K, or 4K. The default is 720p, and higher tiers cost materially more.
Generated audio and scene cuts
Audio is on by default, and Vidu documents direct sound generation and automatic scene cuts; verify the returned task output.
Main parameters
Parameters come from structured model configuration; unsupported or unconfigured fields are omitted.
modelstringSubmodel ID
Default:viduq4-preview / viduq4-previewpromptstringVidu Q4 官方旗舰预览模型，支持单图生视频与多参考生视频，提供 3～16 秒、540p 至 4K 输出、自动分镜和音画同出。
Default:RequireddurationintegerOutput video seconds
Default:3–16squalityenumCurrent billing
Default:540p / 720p / 1080p / 2k / 4kaspect_ratioenumCapability
Default:16:9 / 9:16 / 1:1 / 3:4 / 4:3generate_audiobooleanCapability
Default:trueimage_urlsarrayImage to video: 1 · Reference to video: 1–15
Default:Varies by routeaudio_urlsarrayReference to video: 0–3
Default:Varies by routeCapabilities and usage guidance
Key features
- Image to video · viduq4-preview
- Reference to video · viduq4-preview
- 540p / 720p / 1080p / 2k / 4k output
- 3–16s Duration
- Generated audio output
- Audio-capable workflow
Limits
- Image to video: Images 1
- Reference to video: Images 1–15 · Audio 0–3
- Capability source: https://platform.vidu.cn/docs/api-reference/video-models/vidu-q4-preview/image-to-video
Limits and cautions
- This is a Preview model. Interfaces, prices, and output behavior may change, so production clients should read the live catalog.
- Image-to-video accepts exactly one image. Reference-to-video accepts 1–15 images and 0–3 MP3 files, each 3–12 seconds.
- The vendor limit is 50 MB per file, while an HTTP request body is capped at 20 MB; use public URLs for larger assets.
- Prompts are limited to 20,000 characters and payload to 1,048,576 characters. Persist output links promptly.
- No paid Roko test was authorized, so this page does not claim Roko-verified identity, lip-sync, physics, or visual quality.
API workflow
- Choose image-to-video or reference-to-video; Q4 does not expose text or start/end-frame generation.
- Prepare public image URLs; for reference mode, optionally prepare up to three 3–12 second MP3 files.
- Set model, prompt, duration, quality, and generate_audio. Use aspect_ratio only for reference mode; image mode inherits its input ratio.
- POST /v1/videos/tasks and retain task_id; reuse the same business idempotency key for retries.
- Poll GET /v1/tasks/{task_id} until completed or failed, then persist the result.
Suitable workflows
Character motion
Continue a character into a walk, turn, or camera push; test identity and clothing stability on a short draft first.
Multi-reference brand clips
Combine character, product, and setting images, clearly assigning the role of each reference in the prompt.
Social clips with sound
Generate picture and sound together, then choose 9:16, 1:1, or 16:9 for the destination platform.
How to choose
Choose by documented routes and limits, not an unmeasured quality ranking based on version numbers.
Q4 Preview vs Q3 Pro
Q4 offers image and multi-reference routes up to 4K; Q3 Pro adds text and start/end-frame routes but tops out at 1080p.
Q4 Preview vs Q3 Turbo
Q3 Turbo exposes four routes up to 1080p. Q4 exposes fewer routes but raises image references from 7 to 15 and adds sound references.
Q4 Preview vs Q3 Mix
Both support reference generation; Q4 also adds a single-image route, 540p/2K/4K tiers, and up to three sound references.
Frequently asked questions
What is the Roko model ID?
Use viduq4-preview exactly; do not replace it with a display name or alias.
Which generation modes are supported?
Image-to-video and reference-to-video are supported; text-to-video and start-end-to-video are not.
Which durations and resolutions are available?
Choose 3–16 seconds and 540p, 720p, 1080p, 2K, or 4K. Defaults are 5 seconds and 720p.
How many references can I send?
Image mode requires one image. Reference mode requires 1–15 images and accepts up to three MP3 files.
How do I disable generated audio?
Send generate_audio: false. The fallback estimate then excludes the 15-credit per-task audio add-on.
Why is no video returned immediately?
Generation is asynchronous. Poll /v1/tasks/{task_id} with the returned task ID until completion or failure.
How is the task priced?
Roko pre-authorizes by duration, resolution, and audio, then settles the difference from actual credits × $0.004906.
Why is the page not indexed yet?
The release gate requires four independent cases and two Roko-run cases. Only documentation evidence is currently available.
Sources and evidence scope
Capabilities and limits come from Vidu documentation; Roko routes and prices come from the candidate catalog configuration. Verified 2026-10-13.
- Vidu Q4 Preview image-to-video API Official image input, duration, resolution, audio, and payload limits
- Vidu Q4 Preview reference-to-video API Official image, audio reference, aspect-ratio, and file limits
- Roko API video task docs Submit with /v1/videos/tasks and poll /v1/tasks/{task_id}
- Roko API model catalog Current callable routes, pricing display, and generation configuration
Models in this family
Explore other available variants from the same model family.
Other video models
Explore more models available through the unified asynchronous video task API.
