Vidu Q4 API reference

Pricing (per second)

Billed by resolution for the seconds you request, any whole number from 3 to 16. Both modes cost the same, audio on or off costs the same, and reference images and audio are free. Charged only on success, refunded on failure.

Autenticação

All requests carry the API key in the header:

Authorization: Bearer YOUR_API_KEY

Generate a video

# Image-to-video: one first frame; the clip keeps the image's aspect ratio (prompt optional)
curl -X POST https://api.apimodels.app/v1/video/generations \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "vidu-q4",
    "first_frame_url": "https://example.com/astronaut.png",
    "prompt": "The astronaut waves at the camera, then the camera rises to reveal the station",
    "duration": 5,
    "resolution": "1080p"
  }'

# Reference-to-video: 1-15 reference images (+ up to 3 MP3 voice clips); prompt required
curl -X POST https://api.apimodels.app/v1/video/generations \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "vidu-q4",
    "prompt": "Reference image 1 walks into the cafe from reference image 2, holds up the cup from reference image 3 and says: Morning! Voice from reference audio 1.",
    "reference_image_urls": [
      "https://example.com/character.png",
      "https://example.com/cafe.jpg",
      "https://example.com/cup.png"
    ],
    "reference_audio_urls": ["https://example.com/voice.mp3"],
    "aspect_ratio": "9:16",
    "duration": 8,
    "resolution": "720p"
  }'

# Response (async — a task is created):
# { "code": 200, "msg": "success", "data": { "taskId": "clxxx", "state": "pending" } }

# Poll until completed (result files are kept 30 days)
curl "https://api.apimodels.app/v1/video/generations?task_id=TASK_ID" \
  -H "Authorization: Bearer YOUR_API_KEY"
# { "code": 200, "data": { "state": "completed",
#     "resultUrls": ["https://r2.apimodels.app/videos/clxxx.mp4"] } }

Two modes, one model id

• Image-to-video — one first frame in first_frame_url (image and image_url are accepted too). The clip follows the image's aspect ratio, so aspect_ratio is not accepted (any value other than auto returns 400); prompt is optional; is_rec turns Vidu's prompt enhancement on or off (on by default). Reference audio is not available.
• Reference-to-video — 1-15 images in reference_image_urls (images is accepted too) and a required prompt; add 0-3 MP3 clips of 3-12 seconds each in reference_audio_urls to guide voices; aspect_ratio is 16:9 (default), 9:16, 1:1, 4:3 or 3:4. Name the inputs by position in the prompt: the first image is reference image 1 and the first clip reference audio 1 (Vidu's own examples write them in Chinese as 参考图1 / 参考音频1).

Parâmetros

Notas

• Measured via the API on 2026-10-22 (540p, 3 seconds, one run each): image-to-video took 102 seconds from submission to the finished video, reference-to-video with 2 reference images and 1 MP3 took 173 seconds; output is a 24 fps H.264 MP4 with an AAC audio track and no watermark. Higher resolutions and longer clips were not tested and will likely take longer.
• Asynchronous: create returns a taskId; poll GET /v1/video/generations?task_id= every 5 seconds or so until completed, or pass callback_url. Result files are deleted after 30 days, so download or re-host promptly.
• There is no text-only mode: a request with neither a first frame nor reference images returns 400. A last frame, reference videos and keyframes are not supported, and mixing a first frame with reference images, or sending reference audio with image-to-video, also returns 400. All of these are rejected at create time: no task, no charge.
• Output has no watermark. Vidu Q4 is a Preview, so Vidu may still change how it behaves.
• Content safety checks run upstream; rejected generations are not charged. Only upload faces and voices you have the rights to use.

Try it in the Playground. Capabilities, pricing and a comparison with Seedance 2.5, Kling V3 and FLUX 3 Max are on the Vidu Q4 model page.