ComfyUI-VibeVoice

ComfyUI-VibeVoice

Long-form, multi-speaker conversational TTS and ASR for ComfyUI. Microsoft's VibeVoice, with model download, VRAM management and audio handling handled for you.

Report Bug  ·  Request Feature

[![License: MIT][license-shield]][license-url] [![Last commit][commit-shield]][commit-url] [![ComfyUI][comfy-shield]][comfy-url] [![Model weights on Hugging Face][hf-shield]][hf-url]

[!NOTE] ComfyUI-VibeVoice wraps the Microsoft VibeVoice family — 1.5B / 7B long-form conversational TTS, a 0.5B realtime model, and ASR — as native ComfyUI nodes. Models download on first run into ComfyUI/models/tts/VibeVoice/ and load under ComfyUI's VRAM manager. Nodes appear under WMNodes/sound/tts and WMNodes/sound/asr.

• Features
• Models
• Getting started
• Usage
  • Script format
  • Voice modes
  • Realtime models
  • Node inputs
  • Loading external models
    • Host RAM during an external load
• Performance & stability
  • Feature compatibility & VRAM matrix
  • transformers version support
  • Attention backends
• Changelog
• License
• Acknowledgments
• Star history

❖ Features

Capability	Detail
Multi-speaker TTS	![TTS][task-tts] - Up to 4 distinct voices in one output
Voice cloning	![TTS][task-tts] - Any .wav / .mp3 as a speaker reference
Hybrid scripts	![TTS][task-tts] - Cloned voices plus speakers that reuse a provided reference, in the same script
Speech recognition	![ASR][task-asr] - VibeVoice ASR returns timestamped speaker segments
Quantization	![GGUF][badge-gguf] ![fp8][badge-fp8] ![int8][badge-int8] ![int4][badge-int4] - GGUF, fp8, int8 and 4-bit, plus safetensors
Attention backends	eager / sdpa / flash_attention_2 / sage
Memory	Models download and load under ComfyUI's VRAM manager

⌬ Models

Downloaded on first run into ComfyUI/models/tts/VibeVoice/.

Model	Type	Size	Weights
VibeVoice-1.5B	![TTS][task-tts]	3.0 GB	![][hf-microsoft]
VibeVoice-7B	![TTS][task-tts]	17.4 GB	![][hf-vibevoice]
VibeVoice-Realtime-0.5B	![realtime][task-realtime]	1.5 GB	![][hf-microsoft]
VibeVoice-ASR-HF	![ASR][task-asr]	17.4 GB	![][hf-microsoft]

[!TIP] Quantized weights are optional. Point Load VibeVoice Model at a .gguf or a quantized safetensors file; leave the node unconnected to use the standard model.

⇩ Getting started

ComfyUI Manager — `Mana