Sora 2 Pro

Sora 2 Pro (openai/sora-2-pro) is OpenAI’s flagship video model: cinematic quality, realistic physics and audio. Generation is asynchronous: POST /v1/videos returns a job, then you poll the status and download the mp4.

Parameters

Field	Description
`model`	`openai/sora-2-pro` (or the alias `sora-2-pro`)
`prompt`	Scene description, required
`seconds`	One of 4, 8, 12, 16, 20; defaults to 4
`size`	One of the four sizes below; defaults to `1280x720`
`input_reference`	Image-to-video: a starting frame, `{"image_url": "data:image/...;base64,..."}`
`input_references`	Not supported: the model’s `references` is `false`, a request with the array returns `400`
`seed`	Not supported: the model’s `seed` is `false`; a request with the field returns `400`

The sizes are the exact OpenAI video API dimensions, passed through as is:

Resolution	Landscape	Vertical
720p	`1280x720`	`720x1280`
1080p	`1792x1024`	`1024x1792`

Examples

import requests

job = requests.post(
    "https://api.mixen.ai/v1/videos",
    headers={"Authorization": f"Bearer {API_KEY}"},
    json={
        "model": "openai/sora-2-pro",
        "prompt": "A paper boat sails across a puddle, clouds reflected, macro shot",
        "seconds": "8",
        "size": "1280x720",
    },
).json()
print(job["id"], job["status"])

curl https://api.mixen.ai/v1/videos \
  -H "Authorization: Bearer $MIXEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/sora-2-pro",
    "prompt": "A paper boat sails across a puddle, clouds reflected, macro shot",
    "seconds": "8",
    "size": "1280x720"
  }'

Vertical clip in 1080p

{
  "model": "openai/sora-2-pro",
  "prompt": "Coffee pouring into a glass, slow motion, steam rising in a spiral",
  "seconds": "12",
  "size": "1024x1792"
}

Poll and download

curl https://api.mixen.ai/v1/videos/video_... \
  -H "Authorization: Bearer $MIXEN_API_KEY"

curl -L https://api.mixen.ai/v1/videos/video_.../content \
  -H "Authorization: Bearer $MIXEN_API_KEY" \
  -o sora.mp4

Statuses: queued → in_progress → completed or failed. Cycle details — video guide.

Price

Billing is per second: 720p — 40.1079 ₽ ($0.45) per second, 1080p — 66.8464 ₽ ($0.75). You are charged actual duration × the size’s rate. Current prices — catalog.

Strengths and limits

Cinematic scene quality and realistic motion physics — the pick when the clip is watched, not just embedded.
Audio is present in the finished clip right away — no separate voiceover pass.
Two resolutions up to 1080p, landscape and vertical orientation.
Durations are fixed: only 4, 8, 12, 16 or 20 seconds — no continuous range like Seedance.
Only 16:9 and 9:16: the model does not produce square or ultra-wide frames.
Reference-to-video is not supported; an image can only be passed as the input_reference starting frame.

All models — in the catalog. The general video workflow — video guide.