Skip to main content

Omni Video Models

The Omni video models submit asynchronous tasks via POST /v1/videos and check status with GET /v1/videos/{task_id}.

omni-fast

Text-to-video, first-frame image-to-video, first/last frame, and multi-reference image video.

omni-fast-v2v

Reference-video editing, extension, and regeneration.

Supported Models

ModelDescriptionSupported ModesReference Asset Limits
omni-fastGeneral fast video modelText-to-video, first-frame I2V, first/last frame, multi-imageUp to 5 images; reference video not supported
omni-fast-v2vReference-video modelReference-video editing, extension, regeneration; partial T2V / I2V fieldsReference video up to 15MB

Supported Capabilities

Capabilityomni-fastomni-fast-v2vKey Fields
Text-to-videoYesYesmodel, prompt, seconds, aspect_ratio
First-frame I2VYesYesfirst_image_url
First/last frameYesYesfirst_image_url, last_image_url
Multi-referenceYesYesimages
Reference videoNoYesvideo / video_url / input_video

Unified Endpoint

POST /v1/videos
All Omni video tasks use a JSON request body. On success, an OpenAI-style video object is returned, where id or task_id is used for subsequent queries.

Task Query

GET /v1/videos/{task_id}
Use GET /v1/videos/{task_id} to poll the status. After the task completes, retrieve the generated MP4 from the returned video_url (or data[0].url).

Output Limits

ItemRule
Video durationseconds is recommended as a string, range 4 to 30 seconds
Resolutionresolution example values: 720p, 1080p, 2k, 4k
Aspect ratioaspect_ratio example values: 16:9, 9:16
Reference imagesPublic URL or data:image/...;base64,...; up to 5 images
Reference videoOnly supported by omni-fast-v2v, max 15MB

Sub-API Pages

PageContent
omni-fastText-to-video, image-to-video, first/last frame, multi-image
omni-fast-v2vReference-video editing, extension, regeneration
  1. omni-fast
  2. omni-fast-v2v
  3. Video Model Support Matrix
  4. Task Status API