# Speech-to-Text API

> Whisper tiny.en transcription on Mojo CPU kernels, OpenAI-compatible, $0.003 per audio minute

MCP: https://transcribe.mojojojo.cc/mcp (JSON-RPC over POST). OpenAPI: https://transcribe.mojojojo.cc/openapi.json.
Auth: `Authorization: Bearer mj_live_...`; create a key at https://mojojojo.cc/keys. Usage is billed from mojojojo credits.

## Pricing

- Audio transcribed: $0.003 per minute (10 free per month)

## Endpoints

- `POST /v1/audio/transcriptions` (MCP tool `transcribe_audio`): Transcribe English speech (wav/flac/ogg, base64, up to 300 s) with Whisper tiny.en.

## Limits and measured results

- wav, flac or ogg; mono-mixed and resampled to 16 kHz; up to 300 seconds per request, decoded in 30 second windows.
- English only, smallest Whisper model: expect visibly more errors than larger models. No word error rate is claimed; parity was checked only as identical greedy tokens versus Hugging Face on synthetic clips.
- Speed: 0.93x PyTorch end to end (32 tokens), 0.71x on the encoder.
- Audio is decoded in memory and never written to disk.