ComfyUI-Pocket-TTS 🎙️
January 25, 2026 · View on GitHub
Lightweight CPU-based Text-to-Speech for ComfyUI
Fast, efficient TTS running at 6x real-time on CPU without GPU requirements. Perfect for quick voice generation and prototyping!
✨ Features
- 🚀 Fast: ~200ms latency, 6x real-time on CPU
- 💻 CPU Only: No GPU needed (works on laptops!)
- 🎯 Small Model: Only 100M parameters
- 🎭 8 Built-in Voices: Ready to use
- 🔊 Voice Cloning: Use any audio file
- 📝 Long Text: Handles infinitely long inputs
- ⚡ Low Memory: Uses only 2 CPU cores
📦 Installation
Method 1: ComfyUI Manager (Recommended)
- Open ComfyUI Manager
- Search for "Simple Pocket TTS"
- Click Install
- Restart ComfyUI
Method 2: Manual Install
cd ComfyUI/custom_nodes
git clone https://github.com/ai-joe-git/ComfyUI-Pocket-TTS
cd ComfyUI-Pocket-TTS
pip install -r requirements.txt
🎮 Nodes
1. 🎙️ Pocket TTS (Simple)
All-in-one node for quick TTS
- Input: Text + Voice
- Output: Audio
- Auto-loads model, no setup needed
2. 🎙️ Pocket TTS Model Loader
Load model with specific settings
- Precision: float32/float16/bfloat16
- Outputs model for other nodes
3. 🎙️ Pocket TTS Generate
Generate speech with built-in voices
- 8 voices: alba, marius, javert, jean, fantine, cosette, eponine, azelma
- Fast generation
4. 🎙️ Pocket TTS Voice Clone
Clone voice from audio file
- Upload reference audio to
ComfyUI/input/ - Generates speech in that voice
🎭 Built-in Voices
| Voice | Description |
|---|---|
| alba | Alba Mackenna (default) |
| marius | Marius Pontmercy |
| javert | Inspector Javert |
| jean | Jean Valjean |
| fantine | Fantine |
| cosette | Cosette |
| eponine | Eponine |
| azelma | Azelma |
🔧 Usage Examples
Basic Usage (Simple Node)
Text: "Hello world, this is a test."
Voice: alba
→ Audio Output
Voice Cloning
- Place audio file in
ComfyUI/input/my_voice.wav - Use Pocket TTS Voice Clone node
- Select
my_voice.wav - Enter text
- Get audio in your voice!
Advanced (Model Loader)
[Model Loader] → [Generate] → [Audio Output]
precision: float32
voice: marius
text: "Your text here"
⚡ Performance
Tested on MacBook Air M4:
| Metric | Value |
|---|---|
| Latency | ~200ms first chunk |
| Speed | 6x real-time |
| CPU Cores | 2 cores |
| Model Size | 100M params |
| Memory | ~400MB RAM |
🔄 Workflow Integration
Works with:
- ✅ Video Helper Suite - Save audio
- ✅ Audio Processing Nodes - Effects/mixing
- ✅ Batch Processing - Multiple voices
- ✅ Animation Workflows - Lip sync
🐛 Troubleshooting
❌ "No module named 'pocket_tts'"
# In ComfyUI venv:
pip install pocket-tts
❌ "No audio files found"
Place audio files in:
ComfyUI/input/
Supported: .wav, .mp3, .flac, .ogg
⚠️ Slow generation
- Use
float32precision (fastest on CPU) - Close other applications
- Check CPU isn't throttling
🆚 Comparison
| Feature | Pocket TTS | Other TTS |
|---|---|---|
| Device | CPU only | GPU required |
| Speed | 6x RT | 2-4x RT |
| Latency | 200ms | 500ms+ |
| Model Size | 100M | 1B+ |
| Setup | 1-click | Complex |
📚 Credits
- Pocket TTS: Kyutai Labs
- Paper: Pocket TTS Research
- ComfyUI Node: ai-joe-git
📝 License
MIT License
🚀 Updates
v1.0.0 (2026-01-25)
- Initial release
- 4 nodes
- 8 built-in voices
- Voice cloning support
- CPU-optimized
🤝 Contributing
PRs welcome! Areas for improvement:
- WebAssembly support
- Quantization (int8)
- More voices
- Streaming output
⚠️ Prohibited Use
Voice cloning requires explicit consent. Do not use for:
- ❌ Voice impersonation without consent
- ❌ Misinformation/fake news
- ❌ Harassment or hate speech
- ❌ Privacy violations
See Pocket TTS license for full terms.
Made with ❤️ for the ComfyUI community