Video generation API

Describe a clip in plain words, choose a model, a duration and a frame shape, and get back an MP4 file with a soundtrack. OmevAI plans the shots and camera movement from your description; you choose only what is listed below. The price is known before you start, and you pay only for a clip that is delivered.

Access is enabled on request

The Video API is switched on per key. Write to support@omev.be to enable it. The same models are available in the web cabinet, see Video generation in the cabinet.

Video is asynchronous: create a job, check its status, download the finished file. All calls use the usual host https://llmapi.omev.be and your OmevAI API key.

Models and prices

`model`	Best for	Duration	`resolution`	You can send	Price per second
`omev-video-lite`	simple scenes, low price	4–12 s	`768p`	text; text and a first frame	$0.039
`omev-video`	people, complex motion, camera movement, several actions	4–15 s	`720p` (default), `1080p`	text; text and a first frame	720p $0.069 · 1080p $0.135
`omev-video-premium`	premium quality, clips up to 30 s	4–30 s	`720p` (default), `1080p`	text; text and a first frame; text and up to 5 reference images	text or first frame: 720p $0.075 · 1080p $0.15; with reference images: 720p $0.11 · 1080p $0.22

Frame shapes (aspect_ratio): omev-video-lite takes 16:9, 21:9, 4:3, 1:1, 3:4, 9:16; omev-video and omev-video-premium take 16:9, 4:3, 1:1, 3:4, 9:16.

The price of a clip is the price per second multiplied by the duration you order. An 8-second omev-video clip in 720p costs 8 × $0.069 = $0.552, in 1080p $1.08. An 8-second omev-video-premium clip in 720p costs $0.60 from text or a first frame and $0.88 with reference images.

1. Create a job

curl -sS https://llmapi.omev.be/v1/videos/generations \
  -H "Authorization: Bearer $OMEV_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "omev-video",
    "duration": 8,
    "resolution": "720p",
    "aspect_ratio": "9:16",
    "prompt": "A ceramic mug on a wooden table in morning light, the camera slowly circles the mug",
    "idempotency_key": "order-2026-10-12-0001"
  }'

Field	Required	Values
`model`	yes	`omev-video-lite`, `omev-video` or `omev-video-premium`
`prompt`	yes	the description of the clip, up to 4,000 characters
`duration`	yes	whole seconds within the model's range
`aspect_ratio`	no	a frame shape the model takes, default `16:9`
`resolution`	no	`720p` or `1080p` for `omev-video` and `omev-video-premium`; `omev-video-lite` is always `768p`
`image`	no	a public `https` link to the first frame (all three models); it must open without a login
`references`	no	up to 5 public `https` links to reference images: a character, a product, a logo or a style; `omev-video-premium` only, and not together with `image`
`idempotency_key`	recommended	your own order number, up to 100 characters; repeating it returns the same job and does not charge again

seed, fps, steps, cfg and provider are not accepted (400): OmevAI chooses them.

The answer is 202 with the job:

{
  "id": "vid_0f3c2a9e8b7d4c6a1e5f2b3c",
  "object": "video.generation",
  "model": "omev-video",
  "status": "queued",
  "duration": 8,
  "aspect_ratio": "9:16",
  "resolution": "720p",
  "price_usd": 0.552,
  "access_token": "…"
}

Before the job starts, the price is reserved on the key's balance. When the clip is delivered it is charged; when the generation fails the reservation is released. Keep id and access_token: the token is needed to read the status and download the file. If the answer was lost, send the same request with the same idempotency_key to get the same id and token.

2. Check the status

curl -sS https://llmapi.omev.be/v1/videos/generations/$ID \
  -H "Authorization: Bearer $OMEV_API_KEY" \
  -H "X-Omev-Video-Token: $TOKEN"

Without the token, or with another job's token, the answer is 404. Poll every 10–15 seconds.

`status`	Meaning
`queued`, `preprocessing`	the job is accepted and the description is being turned into a shot plan
`generating`	the clip is being made; `phase` shows `provider_queued` (waiting for a slot), `generating` or `saving`
`processing`	a rare extra check on our side; keep polling
`succeeded`	done: `video.url` is the download link
`failed`	not done: `error.code` and `error.message` say why; the reservation is released

3. Download the clip

curl -sS -L -o clip.mp4 https://llmapi.omev.be/v1/videos/generations/$ID/content \
  -H "Authorization: Bearer $OMEV_API_KEY" \
  -H "X-Omev-Video-Token: $TOKEN"

The answer is the MP4 file with a soundtrack. Download it as soon as the job succeeds and keep it in your own storage.

Full cycle in Python

import os, time, requests

BASE = "https://llmapi.omev.be"
auth = {"Authorization": f"Bearer {os.environ['OMEV_API_KEY']}"}

job = requests.post(f"{BASE}/v1/videos/generations", headers=auth, timeout=60, json={
    "model": "omev-video-lite", "duration": 6, "aspect_ratio": "1:1",
    "prompt": "A cup of hot coffee, steam rising, soft light",
    "idempotency_key": "order-0002"}).json()

read = {**auth, "X-Omev-Video-Token": job["access_token"]}
while True:
    status = requests.get(f"{BASE}/v1/videos/generations/{job['id']}", headers=read, timeout=60).json()
    if status["status"] in ("succeeded", "failed"):
        break
    time.sleep(15)

if status["status"] == "succeeded":
    clip = requests.get(f"{BASE}{status['video']['url']}", headers=read, timeout=600)
    open("clip.mp4", "wb").write(clip.content)
else:
    print(status["error"])

Free lookups

GET /v1/videos/models: the models with their durations, resolutions, frame shapes and prices per second.
GET /v1/videos/price?model=omev-video&duration=8&resolution=1080p: the price of one clip, for example {"price_usd": 1.08, "per_second_usd": 0.135, …}. For omev-video-premium with reference images add &references=1.

Time and limits

A 4-second clip usually takes about 2 minutes with omev-video-lite and omev-video-premium and 10–20 seconds with omev-video. Longer clips take longer.
An account can have up to 2 unfinished video jobs at a time; a third one answers 429.
Regenerating and rating a clip are available in the web cabinet. Through the API, create a new job for a new attempt.
Requests that break the content policy are refused before any charge.

Errors

Code	Meaning	What to do
`400`	a duration outside the model's range, a frame shape or resolution the model does not take, a first frame together with reference images, reference images for a model other than `omev-video-premium`, or a parameter that is not accepted	correct the request using `error.message`
`401`	missing or invalid API key	check the key
`402`	the key's balance is lower than the price of the clip	top up the balance
`403`	video is not enabled for the key	write to support
`404`	unknown job, or a missing or wrong `X-Omev-Video-Token`	check the id and the token
`409`	the `idempotency_key` was already used with other parameters, or the combination is temporarily unavailable	use a new order number; for unavailability, retry later
`429`	2 video jobs are already running	wait until one of them ends
`5xx`	temporary error	retry with the same `idempotency_key` after 10–30 seconds

See Errors and retries for the common format.

Models and prices​

1. Create a job​

2. Check the status​

3. Download the clip​

Full cycle in Python​

Free lookups​

Time and limits​

Errors​