Copilot TTS
A VS Code extension that reads GitHub Copilot Chat responses aloud using a local TTS server.
Supports macOS and Windows 11.
Quick Start
- Run
Copilot TTS: Initialize Copilot TTS once.
- Turn on TTS Chat Mode from the status bar or the
Copilot TTS: Toggle TTS Chat Mode command.
- Use normal Copilot Chat.
- When Copilot finishes a response, it reads the latest assistant reply aloud.
Features
| Feature |
How to use |
| Auto-read normal Copilot chat |
Initialize Copilot TTS, turn on TTS Chat Mode, then use chat normally |
| Replay last spoken response |
Click Play, run copilot-tts.playLast, or use Ctrl+6 on Windows and macOS |
| Stop playback |
Click Stop, run copilot-tts.stopReading, or use Ctrl+Shift+6 on Windows and macOS |
| Change voice |
Voice dropdown in the panel |
| Change speed |
Speed dropdown in the panel or status bar |
Commands
- Initialize Copilot TTS
- Toggle TTS Chat Mode
- Play Last Chat Response
- Stop Reading
- Start TTS Server
- Stop TTS Server
- Clean Local TTS Data
- Toggle Verbose Logging
- Show TTS Server Status
- Set TTS Speed
Configuration
All settings live under copilot-tts.*.
| Setting |
Default |
Description |
pythonPath |
auto-detect |
Optional Python executable override used to run the local TTS server and hook runner. Leave empty to use the managed runtime fallback automatically. |
port |
8765 |
Local TTS server port |
voice |
M1 |
Voice preset used for synthesis |
language |
en |
Synthesis language |
speed |
1 |
Playback speed |
debug |
false |
Enable verbose logging in the Copilot TTS Server output channel. You can also use the Toggle Verbose Logging command. |
readCodeBlocks |
false |
Read fenced code blocks aloud. When false (default) code blocks are silently skipped. |
autoRouteAllChat |
false |
Legacy migration setting used only as a default for restoring chat mode |
Voices
| ID |
Name |
| F1 |
Emma |
| F2 |
Sophia |
| F3 |
Grace |
| F4 |
Luna |
| M1 |
Adam |
| M2 |
Brian |
| M3 |
Charlie |
| M4 |
Daniel |
Supported languages
en ko ja fr de es ru uk pl pt it nl ar hi tr sv da fi cs hu ro bg hr sk sl lt lv et el id vi
How TTS playback works
Responses are played one at a time. If a new response arrives while audio is already playing it is appended to the queue and played automatically when the current item finishes — nothing is interrupted. The panel shows each queued item with its chat name and a playing / queued status.
When you switch to a different chat session the Play button replays that session's last response. Spoken text is cached for the 50 most recently active sessions and survives extension restarts.
When multiple VS Code windows are open, only one window plays audio at a time. The window whose workspace triggered the response always plays it.
Troubleshooting
Play says there is no recent response
The extension replays the last spoken response from its local replay cache. If this is empty, make sure the hook actually fired and inspect the Copilot TTS Server output channel.
Auto-read does not work in normal chat
Run Copilot TTS: Initialize Copilot TTS first, then enable TTS Chat Mode. Normal chat auto-read depends on the Stop hook firing — Initialize Copilot TTS registers the hook file and enables the required VS Code setting chat.useHooks. If that setting is disabled (for example by a workspace settings.json), the hook will never fire regardless of TTS Chat Mode being on.
Server fails to start
Run Copilot TTS: Initialize Copilot TTS again and inspect the Copilot TTS Server output channel.
I want to remove everything before uninstalling
Run Copilot TTS: Clean Local TTS Data before uninstalling. VS Code does not give this extension a reliable uninstall hook, so local cleanup must be triggered explicitly.
License
MIT
