API

BabySea exposes a regional /v1 API for image and video generation. App-level responses use one envelope and one error format across inference providers.

Generations

Create, track, cancel, delete, and list generation records.

Webhooks

Receive signed generation and credit events.

Model schema reference

Browse model identifiers, supported fields, pricing, and request examples.

Base URLs

Region value	Base URL
`us`	`https://api.us.babysea.ai/v1`
`eu`	`https://api.eu.babysea.ai/v1`
`jp`	`https://api.jp.babysea.ai/v1`

API keys are region-specific. A key created in one region cannot authenticate against a different regional hostname.

Authentication

Pass the key in the Authorization header:

Terminal

Authorization: Bearer bye_your_api_key

Keys start with bye_. The full key is shown once when you create or rotate it.

Endpoints

Method	Path	Scope	Description
GET	`/v1/status`	`account:read`	API key and account identity.
GET	`/v1/usage`	`account:read`	Request and generation usage. Accepts `days` from 1 to 90.
GET	`/v1/user/account`	`account:read`	Account profile.
GET	`/v1/user/billing`	`account:read`	Plan, credit balance, and subscription period.
GET	`/v1/estimate/{model_identifier}`	`library:read`	Estimate cost. Accepts `count`, `duration`, `resolution`, and `audio`. `count` defaults to `1` and is capped at `100`.
POST	`/v1/generate/image/{model_identifier}`	`generation:write`	Start an image generation.
POST	`/v1/generate/video/{model_identifier}`	`generation:write`	Start a video generation.
GET	`/v1/content/{generation_id}`	`generation:read`	Retrieve one generation record.
GET	`/v1/content/list`	`generation:read`	List generation records. See generations for pagination defaults.
DELETE	`/v1/content/{generation_id}`	`generation:delete`	Delete a generation and stored output files.
POST	`/v1/content/generation/cancel/{generation_id}`	`generation:delete`	Cancel an in-flight generation when it is still cancelable.
GET	`/v1/health/inference/providers`	`health:read`	Inference provider availability.
GET	`/v1/health/inference/models`	`health:read`	Model availability by inference provider.
GET	`/v1/health/storage`	`health:read`	Storage health.
GET	`/v1/health/cache`	`health:read`	Cache health.
GET	`/v1/library/providers`	`library:read`	Inference provider catalog.
GET	`/v1/library/models`	`library:read`	Model catalog, schemas, inference providers, and pricing.

Generation request fields

Every model has a strict schema. Use /v1/library/models, the playground, or the model schema reference to inspect allowed values for a model.

Field	Type	Notes
`generation_prompt`	string	Prompt text. Required for most models.
`generation_ratio`	string	Aspect ratio, such as `1:1` or `16:9`. Allowed values are model-specific.
`generation_output_format`	string	Output format, such as `png`, `jpg`, `webp`, or `mp4`.
`generation_output_number`	number	Number of requested outputs. Maximum is model-specific.
`generation_input_file`	string[]	Public file URLs for image-to-image or image-to-video models.
`generation_input_file_last_content`	string	Optional last-frame URL for video models that support it.
`generation_duration`	number	Required for video models. Allowed durations are model-specific.
`generation_resolution`	string	Used by models with resolution-specific pricing.
`generation_generate_audio`	boolean	Used by video models with audio pricing.
`generation_provider_order`	string	Optional provider priority. Use `fastest` or a model-supported order such as `replicate, fal`.

Unknown or unsupported fields are rejected for strict model schemas.

Success envelope

Successful app-level API responses use this envelope:

JSON

{
  "status": "success",
  "request_id": "req_...",
  "message": "Operation successful",
  "timestamp": "2026-04-28T12:00:00.000Z",
  "data": {}
}

Paginated responses add total, limit, and offset next to data.

Error envelope

Structured BSE errors use this envelope:

JSON

{
  "status": "error",
  "request_id": "req_...",
  "error": {
    "code": "BSE1002",
    "type": "authentication_failed",
    "message": "Invalid or expired API key.",
    "retryable": false,
    "details": {}
  }
}

See errors for the full code list.

Rate limits

Limits are per account, not per key. All keys under the same account share the same bucket.

Plan	General requests/min	Generation requests/min
Free	30	10
Starter	60	20
Pro	150	50
Scale	300	100
Enterprise	600	200

Rate-limited responses include Retry-After. All rate-limited routes include:

X-RateLimit-Limit
X-RateLimit-Remaining
X-RateLimit-Reset

Concurrency limits

In addition to per-minute rate limits, BabySea caps the number of generations that can be in flight at the same time per account. Image and video are tracked separately.

Plan	Image in flight	Video in flight
Free	2	1
Starter	5	2
Pro	15	5
Scale	30	10
Enterprise	60	20

When the cap is reached, additional generation create requests return BSE1009 (concurrency_limit_exceeded, HTTP 429, retryable). Reduce parallelism or retry after active create requests finish.

Idempotency

POST /v1/generate/image/* and POST /v1/generate/video/* accept an optional Idempotency-Key request header. Use it to safely retry a generation request (for example after a network error) without creating duplicate work or duplicate charges.

Terminal

curl https://api.us.babysea.ai/v1/generate/image/bfl/flux-schnell \
  -H "Authorization: Bearer bye_your_api_key" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: 5b3e1c8a-9c1f-4d2e-9b3f-7c0a1e9d2f4c" \
  -d '{"generation_prompt":"A baby seal playing on Arctic ice"}'

Behavior:

Keys must match [A-Za-z0-9_\-:.]{1,255}. UUIDs work well.
Cached responses are returned for 24 hours after the original request completes.
A replayed response includes the Idempotency-Replayed: true response header. The original request_id, status code, and body are returned exactly as before.
Reusing a key with a different request body returns BSE2015 (idempotency_key_conflict, HTTP 409).
Reusing a key while the original request is still processing returns BSE2016 (idempotency_request_in_progress, HTTP 409, retryable).
Idempotency is scoped per account.

Malformed requests

Requests with an invalid method, path, parameter shape, or body shape can be rejected before BabySea returns a structured BSE error. Validate the request shape before retrying.

Introduction

Get started

Dashboard

Updates

Generations

Webhooks

Model schema reference

Base URLs

Authentication

Endpoints

Generation request fields

Success envelope

Error envelope

Rate limits

Concurrency limits

Idempotency

Malformed requests

Generations

Webhooks

Model schema reference

​Base URLs

​Authentication

​Endpoints

​Generation request fields

​Success envelope

​Error envelope

​Rate limits

​Concurrency limits

​Idempotency

​Malformed requests

Base URLs

Authentication

Endpoints

Generation request fields

Success envelope

Error envelope

Rate limits

Concurrency limits

Idempotency

Malformed requests