Meet the new Scripe, live on October 7.Register

OpenAPI reference · Media

Generate a carousel into the media library asynchronously

POST/media/carousels

Kick off an AI carousel generation: the model picks a design, storyboards the deck, writes every slide, generates any imagery the design calls for, stamps the project's brand identity onto it, and the result is rendered to a PDF by the same pipeline the dashboard's carousel download uses. Returns a Job envelope.

The deck is written in the project's own voice (its brand context and tone-of-voice language) and published under its own identity — author name, handle, headshot and the workspace's logo — exactly as a deck generated in the dashboard is. Nothing about the deck differs by which surface asked for it.

On completion the job's result carries:

  • mediaId — an img_… asset in the project's media library (visible in GET /v1/media, source: GENERATED), whose stored document lets the deck be reopened in the editor;
  • documentKey — the S3 key of the rendered PDF, which you pass straight to PATCH /v1/posts/{postId}/media as { "kind": "document", "document": { "key": "…" } }. That is how a carousel actually reaches LinkedIn;
  • slideCount — how many slides the planner produced;
  • imageSlots — { pending, filled, budgetDeclined, failed }, how the deck's generated-image slots resolved. pending is how many the composed deck asked for; anything not filled shipped with the design's own stand-in art. Check it when you need to know whether the deck you received is complete — a stand-in is visually indistinguishable from a design that wanted no imagery.

Cost: reserved in two phases against the daily AI-spend cap and reconciled with provider actuals after the job runs, so a deck settles at what it really cost. 45¢ is reserved up front for the three chained model calls (design pick, storyboard, slide copy) and the deck's first 20¢ of imagery. The deck's real image-slot count is only known once it has been composed, so each further slot reserves its own cost (5¢ a raster, 11¢ an author cutout) immediately before it is generated. Every reservation is a single atomic conditional update: concurrent calls cannot jointly exceed the cap, and losers receive a 402.

If the cap runs out mid-job the carousel is still delivered — the slots that could not be paid for keep their stand-in art and are counted in imageSlots.budgetDeclined. The cap is never overshot.

Decks typically take two to three minutes, comfortably past the 25 000 ms wait_for_completion_ms ceiling — poll GET /v1/jobs/{jobId} rather than sync-waiting.

Requires the media:write scope.

Authorization

  • Authorizationstringrequired

    Bearer token in the Authorization header.

    Pass Authorization: Bearer scripe_sk_live_<...> (or scripe_sk_test_<...> for test keys) on every request. Keys are scoped to a single workspace and can be revoked from the Scripe dashboard.

    The same header also accepts an OAuth 2.1 access token (scripe_oat_*); both credentials share one scope vocabulary and every operation below documents the scope it requires. An API key can hold every scope named on this surface except webhooks:manage, which is grantable to OAuth tokens only today — the webhook-endpoint operations answer 403 scope_missing to every API key. Operations that name no scope accept any valid token of the workspace.

Header parameters

  • Scripe-Api-Versionstring

    Pin the API version. Format YYYY-MM-DD. Omit to receive the currently rolling default. Unknown versions return 400 version_unsupported.

  • Idempotency-Keystring

    Opaque string (1–64 chars, [A-Za-z0-9_-]) used to dedup retried writes. Within 24h of the first request, the same key

    • same body returns the original response (Idempotent-Replayed: true). Same key + different body returns 409 idempotency_key_conflict.

    Strongly recommended for every write — see /docs/api/v1/idempotency.

Request bodyapplication/json

  • projectIdstringrequired
  • topicstringrequired

    The core idea the deck communicates. A whole post body is fine — the planner reads it as the brief.

  • steerstring

    Explicit instruction that takes priority — audience, angle, tone, things to avoid.

  • outlinestring

    A slide-by-slide outline to honour verbatim. Goes to the PLANNER, the only step that can actually respect slide count and order; omit it to let the planner storyboard the deck.

  • templateIdstring

    Force one of the shipped carousel designs. Omitted — or unknown — and the model picks the best fit for the topic. The workspace's default company design, when one is saved, wins over the model's pick.

  • wait_for_completion_msinteger

    Soft hint — server polls for up to this many ms before returning. Capped at 25 000 ms. A carousel usually takes longer, so polling is the intended path.

Responses

  • 200

    Job enqueued (or finished, on a sync-wait hit).

  • 400

    Malformed request (bad cursor, bad limit, etc.).

  • 401

    Missing, malformed, expired, or revoked API key.

  • 402

    Daily AI-spend cap reached for this workspace. Resets at 00:00 UTC.

  • 403

    Plan not eligible, scope missing, or workspace mismatch.

  • 404

    Resource not found in this workspace.

  • 409

    Idempotency-Key reused with a different body within the 24h dedup window. Pick a fresh key or replay the original body verbatim.

  • 422

    Body shape was JSON but failed validation (unprocessable).

  • 429

    Sliding-window rate limit exceeded.

Example request

bash
curl --request POST \
  --url 'https://api.scripe.io/v1/media/carousels' \
  --header 'Authorization: Bearer <token>' \
  --header 'Content-Type: application/json' \
  --data '{
  "projectId": "proj_a1b2c3d4e5f6g7h8",
  "topic": "string",
  "steer": "string",
  "outline": "string",
  "templateId": "string",
  "wait_for_completion_ms": 0
}'

Example response (200)

json
{
  "data": {
    "id": "job_a1b2c3d4e5f6g7h8",
    "type": "POST_GENERATION",
    "status": "QUEUED",
    "projectId": "string",
    "startedAt": "2026-08-10T09:00:00Z",
    "completedAt": "2026-08-10T09:00:00Z",
    "progress": {
      "progress": 0,
      "message": "string",
      "updatedAt": "2026-08-10T09:00:00Z"
    },
    "result": {},
    "errorCode": "string",
    "errorMessage": "string",
    "attemptCount": 0,
    "estimatedCompletionMs": 0,
    "createdAt": "2026-08-10T09:00:00Z",
    "updatedAt": "2026-08-10T09:00:00Z"
  }
}