Get started › Rate limits

Rate limits

The Munsit API uses a credit-based system with concurrent request limits to ensure fair usage and optimal performance for all users. Here's how the limits work, what happens when you hit one, and how to handle it.

1

Concurrent requests

For streaming text-to-speech requests, Munsit enforces concurrent request limits based on your subscription plan. This ensures stable performance and prevents system overload.

PlanConcurrent request limit
Free / no plan1 concurrent request
Basic2 concurrent requests
Starter5 concurrent requests
Growth10 concurrent requests
Scale20 concurrent requests
EnterpriseUnlimited concurrent requests
2

Live streaming sessions

Streaming sessions on WS /api/v1/listen have their own limits, separate from the concurrent-request limits above.

LimitValue
Concurrent sessions5 simultaneous streaming sessions per API key by default, raised on request. Exceeding it closes the new connection with code 1008.
Idle timeout12 seconds with neither audio nor a KeepAlive message closes the session (code 1011).
PacingAudio may be buffered at most 60 seconds ahead of real time (code 4008).
Frame sizeAt most 1 MiB per binary frame (code 1009).
Audio sent before the session startsAt most 2 MiB or 512 frames while the connection is being authenticated (code 1009).
Session lengthUnlimited while the connection stays active; unbroken speech is force-segmented about every 60 seconds so results keep flowing.
Billing. Seconds of audio received × number of channels, charged from your wallet in 60-second cycles. Connecting requires roughly 60 seconds of wallet runway; running out mid-session closes the connection with code 1008. The closing Metadata event reports the session total as audio_seconds_billed.
3

Request size limits

Each request body, and each WebSocket message, has a size cap. Over the cap, an HTTP request gets 413 with error code 41301 before the body is read, and a WebSocket closes with code 1009.

RequestLimitNotes
JSON request bodies1 MBEvery endpoint, including the upload endpoints when called with JSON (for example with audio_url). 1 MB is 1,048,576 bytes.
File uploads600 MB per requestUpload endpoints (transcription, diarization, denoise, voice cloning, dubbing) accept files as multipart/form-data only; each endpoint's own file limit still applies. Any other body type gets 415 with error code 41501.
WS /listen1 MiB per framePlus 2 MiB or 512 frames of audio sent before the session is authenticated.
WS /websocket/text-to-speech1 MiB per messageText sent before initConnection: 512 messages or 128 KiB.
WS /websocket/speech-to-text (legacy)8 MiB per messageSend 100–500 ms of audio per audio_chunk.
413 · Payload Too Large
{ "errorCode": 41301, "errorMessage": "Request body too large. The limit for this endpoint is 1 MB." }
4

What happens when you exceed a limit

When you exceed your concurrent request limit, the API responds with HTTP 429 and error code 42901 (ConcurrencyLimitError). Either wait for existing requests to complete, or upgrade your plan to increase your limit.

429 · Too Many Requests
{ "errorCode": 42901, "errorMessage": "Concurrency limit exceeded. Maximum 5 concurrent requests allowed. Current: 6. Please upgrade your plan https://app.munsit.com/en/subscription or contact support@munsit.com for more information." }
The full error format — and every other error the API can return — is documented on Errors.
5

Best practices

Three habits keep you clear of 429s in production.

PracticeWhat to do
Monitor your usageCheck your credit balance and concurrent request limits regularly from the dashboard.
Plan aheadConsider your usage patterns when selecting a subscription plan.
Handle rate limitsImplement retry logic with exponential backoff for 429 errors, and queue requests in your application rather than firing them all at once.
6

Subscription plans

For detailed information about subscription plans, credit limits, concurrent request limits, and pricing, visit the subscription page in the dashboard.

Working with an AI assistant? Every page is available as Markdown: add .md to the URL, or send an Accept: text/markdown header. For the whole documentation in one request, point it at llms-full.txt; the page index is llms.txt. Or use Copy Page, top right.