Plans, quotas & limits

The API is plan-gated

API access is a paid feature. Two things follow from that:

POST /v1/keys answers 403 upgrade_required on a plan without it.
Every API-key-authenticated request re-checks the plan, so a downgrade takes effect immediately rather than at the next renewal.

GET /v1/usage is the authority, not this page. Quotas are set per deployment and change with plans; the numbers below describe the hosted backend at the time of writing. Read them from the API rather than hard-coding them.

GET /v1/usage


GET /v1/usage
Authorization: Bearer pm_live_…


{
  "period": "2026-08",
  "renderJobs": 42,
  "renderDurationMs": 3120450,
  "renderedMinutes": 18.4,
  "reservedRenderMinutes": 0.5,
  "apiRequests": 311,
  "assetProcessingBytes": 184320000,
  "limits": {
    "apiEnabled": true,
    "monthlyRenderMinutes": 60,
    "monthlyApiRequests": 1000,
    "maxActiveApiKeys": 5,
    "maxUploadBytes": 104857600
  }
}

Scope: usage:read.

Field	Meaning
`period`	The UTC billing month, `YYYY-MM`. Counters reset when it rolls over.
`renderedMinutes`	Output duration committed by finished renders
`reservedRenderMinutes`	Held by jobs currently queued or processing
`apiRequests`	API-key-authenticated requests this period
`assetProcessingBytes`	Bytes ingested from `image` / `media` inputs
`limits.apiEnabled`	Check this first. `false` means keys will not work.

Gate your integration on limits.apiEnabled, not on the plan name. It is the single flag the server itself uses, and it accounts for a lapsed payment as well as a plan that never included the API.

Quotas on the hosted backend

Quota	API-enabled plan
Render minutes per month	60
API requests per month	1,000
Active API keys	5
Asset ingest per input (by URL)	The smaller of your plan cap (100 MB) and the hard 80 MB fetch limit — so 80 MB
File upload (`POST /v1/assets`)	Your plan’s `maxUploadBytes` — 100 MB

A quota breach is 403 with a machine-readable code, not a silent truncation:


{
  "code": "quota_exceeded",
  "kind": "render_minutes",
  "limit": 60,
  "used": 59.5,
  "message": "Monthly render minutes quota reached."
}

How render minutes are counted

Minutes are reserved before the render and settled after it — the reserve is the template’s durationSeconds, converted to minutes.

This is why a job can be refused while renderedMinutes still looks comfortable: the check is renderedMinutes + reservedRenderMinutes + this job. Reserving up front is what stops two concurrent requests from both passing the same last-minute check and jointly overrunning the quota.

The reservation is released when the job reaches any terminal state, including cancelled and failed. A cancelled render costs you nothing.

Note that the reserve is based on the template’s output duration, not on how long the render takes. A 6-second clip that takes four minutes of CPU to produce still costs 0.1 minutes of quota.

Request counting

Only API-key requests are counted against monthlyApiRequests. Editor and dashboard traffic on a session JWT is not. Every /v1 call with a key counts — including the polling loop, which is a reason to prefer webhooks on a tight quota.

Uploaded and ingested bytes cost the same

assetProcessingBytes meters both routes to the same outcome: a file sent to POST /v1/assets, and an asset fetched from a URL you passed as an input. There is no cheaper path, deliberately — the choice between them should be about what your workflow is holding, not about billing.


Lifetime	24 hours, then deleted
Counted against	`assetProcessingBytes`
Size cap	`limits.maxUploadBytes` — over it is 413 naming the limit
Types	PNG, JPEG, WebP, GIF, SVG, MP4, WebM, QuickTime, MP3, WAV, M4A

Uploads are temporary because they exist to feed a render. For a permanent asset library, use the editor — /assets is a separate, session-authenticated surface and is not part of this API.

Importing a project

POST /v1/templates/import costs one API request and nothing else — no render minutes, no asset bytes. The bundle’s own asset blobs are never ingested, because the render service cannot use bytes that live on your disk; that is what the derived inputs are for.

The template limits below still apply to what the imported composition declares: a project at 8192px or a 90-minute composition is refused.

Rate limits

Separate from the monthly quota, and evaluated per account:

Limit	Value	Answer
Render creation	60 per minute	429 `rate_limit_exceeded`
All requests	300 per minute per IP	429

Hard limits

These are not plan-dependent — they apply to every deployment.

Limit	Value
Output format	MP4 (H.264) only
Output dimensions	64–4096 px per side, at most 16 megapixels
Output fps	1–120, integer
Template `width` / `height`	≤ 8192
Template `fps`	≤ 240
Template `durationSeconds`	≤ 3600
Request body	10 MB
Asset fetch	80 MB
Template `name`	120 characters
Template `description`	500 characters
Page size on any list	100
`Idempotency-Key`	8–128 URL-safe characters

Not supported


Alpha in output	MP4 has none — transparency flattens onto the comp background
Audio	Not muxed by the hosted renderer yet
Non-MP4 formats	WebM, ProRes, PNG sequence, Lottie, SVG and GIF are desktop-export only
Plugin effects	Skipped silently — the hosted renderer has no plugins installed
Layer or keyframe CRUD	Not exposed; the API renders templates, it does not edit them
Reading a template back as a `.motion`	Import is one-way; there is no export endpoint