AI Content Drop REST API Reference

Base URL: https://aicontentdrop.com/v1 · Spec: /openapi.json · Format: JSON only, never HTML, on every status code.

Endpoints

Method and path	Key	What it does
`GET /v1`	no	Machine-readable index of this API
`GET /v1/models`	no	Every model with its flat credit cost
`GET /v1/models/{id}/cost?quantity=N`	no	Quote one model before committing
`POST /v1/models/cost`	no	Quote up to 50 model/quantity pairs in one call
`POST /v1/batch`	no	Run several read operations in one request
`GET /v1/me`	yes	Account, plan, and credit balance
`POST /v1/generate/video`	yes	Start a video generation. Returns 202 + `Location`
`POST /v1/generate/image`	yes	Same contract for images
`GET /v1/videos/{id}`	yes	Poll one generation
`GET /v1/videos?limit=&cursor=`	yes	Cursor-paginated generation history

Generating

POST /v1/generate/video accepts:

Field	Type	Notes
`prompt`	string	Required. What the video should show.
`aiModel`	string	Model id from the catalogue. Omit to let the platform choose.
`duration`	number	Seconds. Defaults to 5.
`aspectRatio`	string	`16:9`, `9:16`, or `1:1`. Defaults to `16:9`.
`imageUrl`	string	Public image URL to animate, for image-to-video.

Three request headers change the contract, and all three are opt-in:

Idempotency-Key: <uuid> — a retry with the same key returns the FIRST response instead of starting a second generation. Reusing a key with a different body is a 409, not a silent replay. Keys live 24 hours.
X-Sandbox: true — returns a synthetic completed job without calling a provider or charging anything. See sandbox mode.
Accept: application/json — the default and the only representation.

The success response is 202 Accepted, not 200, because the render has not happened yet:


{
  "video": { "id": "…", "status": "generating", "aiModel": "kling_3_0", "creditsUsed": 0 },
  "poll_url": "https://aicontentdrop.com/v1/videos/…",
  "poll_interval_seconds": 5
}

Location carries the same poll URL and Retry-After: 5 carries the interval. Poll until status is completed or failed.

Paginating history

GET /v1/videos takes limit (1–100) and an opaque cursor. The response carries next_cursor; pass it back to continue. offset still works and is deprecated — it is announced in the spec with a Deprecation header and will keep working for at least 180 days.

Model ids

Ids use underscores (veo_3_fast). Dashes are accepted on input and normalised. A few marketplace aliases resolve to their provider id. If a model "does not exist", check the id form before anything else.

Errors

Every failure is JSON with a machine code. The full table is in error codes.


{ "error": { "code": "insufficient_credits", "message": "This generation needs 22 credits; the account holds 4." } }