VS Code Voice Assistant
Speak naturally and turn your voice into raw text without leaving VS Code.
VS Code Voice Assistant is a privacy-first voice-to-text extension powered by
whisper.cpp. Recording and transcription happen locally on your computer—no
account, API key, backend, or audio upload is required.
Features
- Record from your default microphone inside VS Code
- Transcribe speech locally with Whisper
- Access a compact interface in the right-side Secondary Side Bar
- See audio-reactive feedback while speaking
- Continue recording to append to your current transcript
- Optionally insert the latest transcript into the active VS Code terminal
- Copy the complete transcript to your clipboard
- Reset the transcript with one click
- Automatically download and verify the required runtime and model on first use
Getting Started
- Open Voice Assistant from the microphone icon in the status bar.
- Select the large microphone button and begin speaking.
- Select the pause button to stop recording and start transcription.
- Copy the transcript, continue recording to add more text, or enable
Insert into active terminal to place the latest transcript in your
current terminal input.
Terminal insertion is opt-in and does not press Enter automatically. You can
review or edit the inserted text before submitting it to Git Bash, PowerShell,
CMD, Codex CLI, Claude CLI, or any other active terminal workflow.
The first recording downloads the Windows whisper.cpp runtime and the English
base.en model (approximately 148 MB). These files are verified, stored in VS
Code's extension storage, and reused for future offline transcription.
Privacy
- Audio and transcription stay on your machine.
- Temporary recordings are deleted after transcription.
- No telemetry, authentication, cloud storage, or audio upload is used.
Requirements
- Windows x64
- VS Code 1.96.2 or newer
- A working default microphone
- Internet access for the first-use runtime and model download
macOS and Linux support is planned for future releases.
Current Scope
This release provides raw English voice-to-text with optional active-terminal
insertion. AI cleanup, filler-word removal, editor insertion, multilingual
transcription, and Ollama integration are planned for later versions.
Found an issue or have an idea? Visit the
GitHub repository.
