Back to Knowledge Base

Molten Code Voice Dictation Setup

Run Molten Code with Faster Whisper so you can dictate prompts from the Code UI.

Use this guide to run Molten Code with local voice dictation through a Faster Whisper service.

Overview

This Docker Compose setup runs two services:

  • agent: the Molten Code application.
  • whisper: the Faster Whisper transcription service.

The agent connects to the whisper container for local speech processing.

docker-compose.yml

View the full docker-compose.yml source file.

Start the Stack

  1. Download the linked source file and save it as docker-compose.yml.
  2. Run docker compose up -d from the same directory.
  3. Open http://localhost:3331.

Use Voice Dictation

Once the stack is running, use the voice controls in the Molten Code prompt composer.

  1. Click the voice dictation button in the composer.
    Molten Code voice dictation button in the prompt composer.
    Click the dictation button to start voice input.
  2. Speak while Molten Code shows the capturing state.
    Molten Code composer showing the voice capture state while listening.
    The capture state confirms the UI is listening for speech.
  3. Review the transcribed text before you send the prompt.
    Molten Code prompt composer showing dictated text after transcription.
    Whisper returns the transcription to the composer for review.

What This Does

  • Starts Molten Code in the agent container.
  • Starts Faster Whisper in the whisper container.
  • Connects the agent to Whisper for speech transcription.
  • Exposes the Molten Code UI on port 3331.