Skip to content

STT Configs

Use the STT tab to configure a whisper.cpp HTTP server connection for speech recognition.

Only one STT backend is expected. It is shared by every simulation rather than assigned per world or per component.

Fields

Field Description
connection A whispercpp connection whose base URL points to a reachable whisper.cpp server.
model Optional model label for display or backend selection where supported.
language Spoken language code, such as en, zh, or auto.
translate Whether to translate the transcription into English.
temperature Decoding temperature.
temperature_inc Temperature increment used on decoding fallback.
initial_prompt Optional prompt text to bias vocabulary or context.
carry_initial_prompt Whether to prepend the initial prompt for each request.

When the backend runs in Docker, make sure the whisper.cpp URL is reachable from the backend container, not just from the host browser.