Use voice input with speech-to-text in OpenRails Chat
OpenRails integrates a speech-to-text engine for audio transcription, allowing you to use voice input in conversations. The transcription engine can run locally for privacy or remotely via a cloud API for convenience. Transcribed text is automatically inserted as a chat message.
| Mode | Description | Requirements |
|---|---|---|
| Local | Runs the speech-to-text engine on your server hardware. Audio never leaves your network. | Transcription model downloaded locally, sufficient GPU/CPU resources |
| Remote | Sends audio to a cloud speech-to-text API for transcription. | Valid OpenAI API key configured in LLM Keys |
Navigate to a bot and open the chat interface.
Click the microphone icon next to the message input field. Your browser will request microphone permission if not already granted.
Speak clearly into your microphone. A visual indicator shows that recording is in progress.
Click the microphone button again or press Escape to stop recording. The audio is sent for transcription.
The transcribed text appears in the message input field. Review the transcription, make any edits, and press Enter to send.
The following audio formats are supported for file upload transcription:
Beyond chat voice input, the speech-to-text engine is also used in the document ingestion pipeline to transcribe uploaded audio and video files. When you upload an audio file to a data lake, it is automatically transcribed, and the transcript is chunked and indexed like any other document.