VOQR — Voice for AI Chat
Talk to any AI. Hear every answer. Local. Private. Yours.
(rhymes with "poker")
Click to talk. AI responds by voice. All local. Watch with audio →
VOQR adds voice interaction to every AI chat provider in VS Code. Speech-to-text and text-to-speech run entirely on your machine — your voice data never leaves your computer.
Features
- Universal — works with Copilot, HuggingFace, Cerebras, AI Toolkit, and any OpenAI-compatible provider
- Local & private — STT (Whisper) and TTS (Kokoro) run on your hardware. No cloud. No telemetry
- Streaming responses — hear AI responses sentence-by-sentence as they generate, not after a long wait
- Push-to-talk or hands-free — toggle between manual and voice-activated modes
- Smart text normalization — AI responses full of markdown, code blocks, and URLs are cleaned up for natural speech
- Setup wizard — guided onboarding walks you through installing and configuring your first AI provider
- Customizable — adjust speech speed, mute/unmute TTS, choose your AI model
Quick Start
- Install VOQR from the VS Code Marketplace
- Click the mic button in the status bar (or use your platform's keybinding — see Keybindings)
- Pick an AI provider — the setup wizard will guide you if you don't have one installed
- Talk — VOQR transcribes your speech, sends it to your AI, and reads the response aloud
That's it. No API keys for VOQR, no accounts, no configuration required.
Supported Providers
VOQR works with any AI provider that implements the VS Code Language Model API:
| Provider |
Cost |
Setup |
| GitHub Copilot |
Copilot subscription |
Install Copilot extension |
| HuggingFace |
Free tier available |
Install extension + API token |
| Cerebras |
Free tier available |
Install extension + API key |
| AI Toolkit |
Free (local models) |
Install extension |
| OpenAI Compatible |
Varies |
Install extension + endpoint config |
The built-in setup wizard walks you through each provider step by step.
Requirements
| Component |
Details |
| VS Code |
1.99.0 or later |
| OS |
Linux, Windows, macOS |
| STT |
whisper.cpp (auto-detected, or specify path in settings) |
| TTS |
Python 3.10–3.13 + Kokoro (auto-installed on first use) |
VOQR auto-detects whisper.cpp and Python on your system. If they're not found, you'll be guided through setup.
Keybindings
| Action |
Linux |
Windows |
macOS |
| Push to Talk |
Ctrl+Shift+Space |
Ctrl+F9 |
Cmd+Shift+Space |
| Toggle Voice Panel |
Ctrl+Shift+V |
Ctrl+Shift+V |
Cmd+Shift+V |
The status bar mic button is always available as an alternative — no keybinding needed.
Settings
| Setting |
Default |
Description |
voqr.inputMode |
pushToTalk |
pushToTalk or voiceActivity (hands-free) |
voqr.ttsSpeed |
1.0 |
Speech speed (0.5× – 2.0×) |
voqr.ttsAutoPlay |
true |
Automatically speak AI responses |
voqr.sttBackend |
auto |
auto (whisper.cpp), faster-whisper (GPU), or external |
voqr.sttModelSize |
tiny |
Whisper model: tiny, base, small, medium |
voqr.debug |
false |
Verbose logging to output channel |
Quick controls for speed and mute are also available directly in the VOQR panel.
Privacy
VOQR processes all voice data locally:
- Speech-to-text: whisper.cpp runs on your CPU — audio never leaves your machine
- Text-to-speech: Kokoro TTS runs on your CPU — no cloud synthesis
- No telemetry: VOQR collects zero usage data
- No account required: Free tier works without sign-up
For a voice extension, "we collect no data" is a feature, not a gap.
How It Works
You speak → Mic capture → Voice activity detection (Silero VAD)
→ Speech-to-text (whisper.cpp) → AI model (your chosen provider)
→ Streaming response → Text normalization → Text-to-speech (Kokoro)
→ You hear the answer — sentence by sentence, as it generates
Comparison
|
VOQR |
VS Code Speech |
Wispr Flow |
Superwhisper |
| Local STT |
Yes |
Yes |
No (cloud) |
Yes |
| Local TTS |
Yes |
Limited |
No |
No |
| Any AI provider |
Yes |
Copilot only |
Any text field |
Any text field |
| VS Code native |
Yes |
Yes |
No |
No |
| Custom voice |
Yes |
No |
No |
No |
| Offline capable |
Yes |
Yes |
No |
STT only |
| Price |
Free |
Free |
$15/mo |
$8.49/mo |
Roadmap
See ROADMAP.md for the full plan. Highlights:
- v0.4 — Zero-config: bundled STT and TTS, no external dependencies
- v1.0 — Pro features: speaker verification, custom voice blending, voice commands for VS Code, audio + text sync with pause/resume
Contributing
VOQR is MIT licensed. Issues, feature requests, and voice quality reports welcome:
License
MIT — see LICENSE.
Open-source attributions: ATTRIBUTIONS.md.
VOQR is an independent project and is not affiliated with, endorsed by, or sponsored by any AI provider. All product names, trademarks, and registered trademarks are property of their respective owners.
