Skip to main content

Video generation models

Openference exposes video models on POST /v1/videos with polling via GET /v1/videos/. Use the model ids below in your request body. Aliases are also accepted (e.g. agnes-video-2.5-flashAgnes-Video-2.5-Flash).

Agnes Video 2.5 Flash (Agnes-Video-2.5-Flash)

Task: text-to-video, keyframe control, image/audio reference
Output: 720P MP4, 4–12 seconds

Text-to-video example

Save video_id from the response, then poll:

Errors

See POST /v1/videos and Common errors for the full list.