Rate Limits
API access requires Pro credits — available pay-as-you-go or via subscription. Signing up free grants 500 Pro credits immediately.
Concurrency limits
Each account can have up to 18 jobs queued + running at a time. Submitting beyond this limit returns 429 with a Retry-After header.
Up to 6 of those jobs will run concurrently — additional jobs stay queued until a running slot opens. Creating multiple API keys does not increase these limits; they are per account.
Response headers
| Header | When | Value |
|---|---|---|
Retry-After | 429 response | Seconds to wait before retrying |
Polling rate limit
Each job can be polled at a sustained rate of 1 request per second per API key. A token bucket allows bursting up to 30 requests — after which requests are limited to 1/second until the bucket refills. If you exceed this, the API returns 429 with a Retry-After header.
Most conversions complete in under 10 seconds — the recommended intervals below keep you well within limits.
Recommended polling intervals
| Conversion type | Poll interval |
|---|---|
| Images (HEIC, PNG, JPG, WebP, AVIF) | 1-2 seconds |
| Documents (PDF, DOCX, PPTX) | 3-5 seconds |
| Audio transcription | 5-10 seconds |
| Large files (> 50 MB) | 5-10 seconds |
Best practices
- Respect
Retry-After— wait the indicated seconds before retrying. - Exponential backoff — if no
Retry-Afterheader, back off: 1s, 2s, 4s, 8s, max 30s. - Poll, don't spin — use the intervals above instead of tight loops.