Cartesia Ink vs Mistral Voxtral Transcribe

Cartesia Ink

Ink 2 streams transcripts with turn detection built into the model, documented by an AsyncAPI file with typed ranges and structured errors. The terms, last revised 14 June 2024, let Cartesia train on inputs and outputs unless the customer opts out, and zero data retention is an Enterprise setting. Ink 2 has no batch endpoint and no diarisation.

Mistral Voxtral Transcribe

Batch transcription costs $0.003 a minute and takes one multipart call with a file, a URL or an uploaded file ID. Rate limit numbers are shown only in the console, and timestamps cannot be combined with a set language, nor diarisation with the realtime model.

Cartesia Ink

1. Use wss://api.cartesia.ai/stt/turns/websocket with model=ink-2, encoding, sample_rate and cartesia_version=2026-08-14. All four are required.
2. Send raw mono audio in chunks of about 100 ms at the speed it was spoken. Pushing a whole file into the socket can return an internal server error.
3. Read the final text from turn.end only. transcript is cumulative within a turn, so joining turn.update events duplicates text.
4. Send {"type": "close"} after the last audio and keep reading until the server closes the socket, or the buffered tail is lost.
5. Check encoding and sample_rate against the source before sending. The docs say the server might not return an error when they are wrong.

Mistral Voxtral Transcribe

1. Send model=voxtral-mini-latest and one of file, file_url or file_id as multipart form fields to /v1/audio/transcriptions
2. Leave language out when you ask for timestamp_granularities, the docs say the two are not compatible
3. Use the batch endpoint for diarisation. voxtral-mini-transcribe-realtime-2602 does not accept diarize
4. Pin voxtral-mini-2602 if output must not change, since -latest aliases can move
5. Mint browser tokens with POST /v1/client/sessions close to connection time and pass them in Sec-WebSocket-Protocol, never the API key