v1.2.2 - 4-Bit Quantization Support

[Enemyx-net]

🚀 4-Bit Quantized Model

• New Model Option: Added VibeVoice-Large-Quant-4Bit to the model selection menu (Thanks to DevParker for quantizing the model)
• Memory Optimization: Uses only ~7GB VRAM instead of ~17GB for the full model
• Performance: Maintains excellent quality while significantly reducing memory requirements

💾 Memory Savings

VRAM Usage Comparison

• VibeVoice-Large: ~17GB VRAM
• VibeVoice-Large-Quant-4Bit: ~7GB VRAM
• Savings: 60% reduction in memory usage

✨ Benefits

• ✅ Accessible to More GPUs: Run on cards with 8GB+ VRAM
• ✅ Faster Loading: Reduced model size means quicker initialization
• ✅ Batch Processing: Fit larger batches in the same VRAM budget
• ✅ Quality Preserved: Minimal quality loss with 4-bit quantization

⚙️ Requirements

For 4-Bit Model:

• GPU: NVIDIA GPU with CUDA support
• Library: bitsandbytes must be installed

  pip install bitsandbytes

• VRAM: Minimum 8GB recommended

💾 Installation

Install via ComfyUI Manager or manually:

git clone https://github.com/Enemyx-net/VibeVoice-ComfyUI

📋 Usage

• Ensure bitsandbytes is installed
• Select VibeVoice-Large-Quant-4Bit from the model dropdown
• Enjoy the same great quality with lower VRAM usage

🔧 Notes

• The 4-bit model is only available for NVIDIA GPUs with CUDA
• First-time use will download the quantized model (~7GB)

---

This discussion was created from the release v1.2.2 - 4-Bit Quantization Support.