Video models

Twelve models are served on POST /v1/videos. They all take the same request schema — what differs is clip length, frame size, accepted aspect ratios and price per second. Pick one from the matrix, then pass its ID as the model field.

Capability matrix

Modes — t2v text-to-video, i2v image-to-video, v2v video-to-video. You never send a mode; it is inferred from whether the request carries frame_images or input_references. See Generation mode.

Ratio sets — each set contains the one above it:

Audio — all twelve models render a synchronized soundtrack; generate_audio defaults to true.

The video_attributes object

Every video model in the Models API response carries a video_attributes object. It is the authoritative description of what that model accepts — the matrix above is a snapshot of it.

{
  "modes": ["t2v", "i2v", "v2v"],
  "resolutions": ["480p", "720p", "1080p"],
  "default_resolution": "720p",
  "min_duration_seconds": 4,
  "max_duration_seconds": 30,
  "supports_audio": true,
  "aspect_ratios": ["21:9", "16:9", "4:3", "1:1", "3:4", "9:16", "adaptive"]
}

Query the live list

The Models API is public and takes no API key. Filter on supported_endpoints to keep only video models:

curl -s https://api.ofox.ai/v1/models \
  | jq -r '.data[] | select(.supported_endpoints[]? == "/v1/videos") | .id'

import requests
 
models = requests.get("https://api.ofox.ai/v1/models").json()["data"]
 
for m in models:
    if "/v1/videos" in m.get("supported_endpoints", []):
        va = m["video_attributes"]
        print(m["id"], va["modes"], va["resolutions"], va["max_duration_seconds"])

const resp = await fetch('https://api.ofox.ai/v1/models')
const { data } = await resp.json()
 
for (const m of data) {
  if (!m.supported_endpoints?.includes('/v1/videos')) continue
  const va = m.video_attributes
  console.log(m.id, va.modes, va.resolutions, va.max_duration_seconds)
}

Pick by requirement

Next