Skip to content
TomoriBot

Speech-to-Text (Local Engines)

Transcribe voice messages and audio attachments locally so TomoriBot can understand spoken audio in chat without sending recordings to third-party cloud APIs.

TomoriBot treats speech-to-text as the transcription custom endpoint capability. Audio attachments are transcribed in the background and added to conversation context when an active transcription endpoint exists.

Visible transcript posting is separate. /config > Behavior > Notice Behavior > Speech Transcripts controls whether transcripts appear in chat. Turning it off keeps incoming transcripts in conversation context. The notice visibility controls affect failure notices separately.

Available engines

WhisperXBeginner-friendly local transcription using the reference WhisperX server under servers/stt.
whisper.cppUse a lightweight whisper.cpp HTTP server when it exposes an OpenAI-compatible transcription endpoint.
KoboldCPPUse KoboldCPP Whisper-based STT when your build exposes OpenAI-compatible audio transcription.

Setup flow

  1. Start one STT server from the engine guides above.
  2. In /providers, choose + Add New Custom Endpoint (or Add New Custom Endpoint) and select openai-compatible-transcription (OpenAI Compatible Transcription).
  3. Select the saved endpoint and use its model dropdown to add the model code reported by the server.
  4. Open /config > Models > Switch Models and select the registered endpoint under Speech to Text.

The transcription selection is independent of TTS (Voice Design) and TTS (Voice Clone). Adding Google or OpenRouter speech models does not configure incoming transcription.

Transcript flow

When a transcription endpoint is active, TomoriBot reads supported audio attachments in the background and adds the transcript to chat context. No persona assignment is needed.

Enable Speech Transcripts in /config > Behavior > Notice Behavior if you also want transcripts posted visibly in chat.

ElevenLabs users should choose Add New Provider and ElevenLabs in /providers; it registers transcription alongside speech.