Video generation API

Video generation, behind one HTTP API.

OmniPort runs the generation pipeline and bills it by the second. Call it through the OpenAI Videos API your SDK already speaks, or through native endpoints that take a resolution tier directly.

Read the docs See pricing

Drop-in OpenAI SDK
Prepaid credits on your account
Failed jobs refunded automatically

POST /api/v1/videos

curl https://www.omniportapi.com/api/v1/videos \
  -H "Authorization: Bearer $OMNIPORT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "sora-2",
    "prompt": "a paper boat drifting through a neon city",
    "size": "720x1280",
    "seconds": 8
  }'

# 200 OK
{ "id": "video_9f2c…", "object": "video", "status": "queued" }

Two ways to call it

Same account, same credit balance, same task ownership. Pick the shape that fits your client — you can mix both from the same key.

OpenAI-compatible/api/v1/videos

Point the OpenAI SDK at this host and keep your existing code. Model names, size and seconds are translated to the pipeline for you.

POST /api/v1/videostext to video, or with a reference image
GET /api/v1/videos/{id}status, progress, errors
GET /api/v1/videos/{id}/contentthe finished MP4

Billed per model + secondEndpoints

Nativegenerations

Write against the generation scenarios directly: first / last frame, reference to video, and upscale. You choose the resolution tier; the request goes upstream as-is.

POST /api/v1/fl2v/generationsfirst frame, optional last frame
POST /api/v1/r2v/generationsimages, video and audio references
POST /api/v1/upscale/generationsre-render an existing video at a higher tier

Billed per resolution + secondEndpoints

Every endpoint

All paths are relative to this host and authenticated with a single bearer key. There is no versioned SDK to install.

OpenAI-compatible surface
Method	Path	Description
POST	/api/v1/videos	Create a video job. JSON, or multipart with an input_reference file.
GET	/api/v1/videos	List your videos. limit / after / order.
GET	/api/v1/videos/{id}	Fetch one video with its status and progress.
GET	/api/v1/videos/{id}/content	Redirect to the finished MP4.
DELETE	/api/v1/videos/{id}	Delete a video. Queued jobs are cancelled and refunded.
GET	/api/v1/models	Models enabled on this deployment.

Native surface
Method	Path	Description
POST	/api/v1/fl2v/generations	First / last frame to video. Up to 2 images, prompt required.
POST	/api/v1/r2v/generations	Reference to video. Up to 8 images, 3 videos, 3 audio clips.
POST	/api/v1/upscale/generations	Upscale an existing video. Exactly one videoUrls entry.
GET	/api/v1/generations	List your tasks. page_num / page_size.
GET	/api/v1/generations/{id}	Fetch one task. Finished tasks include artifacts.

Pricing

Priced per second of generated video, with a minimum charge per job. Prices are in credits (1 credit = 1 US cent), drawn from your prepaid balance — no invoices, no rounding surprises.

Per model — OpenAI-compatible surface
Model	Credits / second	Minimum credits / job
sora-2	2	10
sora-2-pro	4	20

Per resolution — native surface
Tier	Typical use	Credits / second	Minimum credits / job
480P	Drafts and prompt iteration	1	5
768P	Everyday output	2	10
1080P	High detail	3	15
2K	Final delivery	5	25

How billing works

The full amount in credits is held when the job is created. If your credit balance does not cover it, the request fails with 402 insufficient_quota and nothing is queued.
The pipeline renders at least 5 seconds. A request for seconds: 4 is accepted and billed as 5 seconds.
Failed or cancelled jobs are refunded automatically, once per job.
Upscale is billed as a single 5-second unit; its real duration follows the source video.
Top up in the console — card and PayPal, credited to your account and shared by its keys.
Prices above are platform defaults; this deployment's price list has not been configured yet.

Buy credits in packs

Packs are priced in US dollars and credited as credits (1 credit = 1 US cent) on your account. Every job draws from that balance. Buy once — no subscription, and nothing expires.

Prepaid packs
Pack	You pay	You receive	Bonus	Best for
Starter	$10.00	1,000 credits	—	Enough for a few short clips — good for wiring up the API.
Standard	$50.00	5,500 credits	+10%	10% bonus credit. Where most teams land after the first test.
Pro	$200.00	24,000 credits	+20%	20% bonus credit. For continuous rendering across projects.

Buy credits How credits work

How a job runs

Generation takes minutes, so every call is asynchronous: submit, poll, download.

CreateSend the prompt and shape. Parameters are validated and the charge is held before anything is queued.
PollRead the job whenever you like. Every read syncs the live status from the pipeline — there is no stale cache to work around.
DeliverFinished jobs point at an MP4 (/content, or artifacts on the native surface). Failures refund themselves.

Auth and errors

One header, one error shape. If your client can parse a JSON error, it can handle every failure this API produces.

Authentication

Authorization: Bearer sk-op-...

Keys are issued by the OmniPort operator. A key spends the credit balance of the account it belongs to, and only sees its own jobs.

Error envelope

{
  "error": {
    "message": "resolution must be one of 480P, 768P, 1080P, 2K.",
    "type": "invalid_request_error",
    "param": "resolution",
    "code": "invalid_value"
  }
}

400 invalid parameter · 401 bad key · 402 out of balance · 404 unknown job · 429 upstream busy.