Nodes/VibeVoice ComfyUI/VibeVoice Multiple Speakers
ComfyUI Node

VibeVoice Multiple Speakers

Generate multi-speaker conversations with up to 4 distinct voices using Microsoft VibeVoice

By Enemyx-net·Created 11 months ago·Updated 6 months ago· 1,518
VibeVoice Multiple Speakers
  • speaker1_voice
  • speaker2_voice
  • speaker3_voice
  • speaker4_voice
  • lora
  • audio
text[1]: Hello, this is the first speaker. [2]: Hi there, I'm the second speaker. [1]: Nice to meet you! [2]: Nice to meet you too!
modelNo models found
attention_typeauto
quantize_llmfull precision
free_memory_after_generatetrue
diffusion_steps20
seed42
cfg_scale1.30
use_samplingfalse
temperature0.95
top_p0.95
voice_speed_factor1.00
CategoryVibeVoiceWrapper

Inputs (17)

NameTypeDefaultDescription
textSTRING[1]: Hello, this is the first speaker. [2]: Hi there, I'm the second speaker. [1]: Nice to meet you! [2]: Nice to meet you too!Text with speaker labels. Use '[N]:' format where N is 1-4. Gets disabled when connected to another node.
modelCOMBONo models foundSelect a model from ComfyUI/models/vibevoice/ folder. Large is recommended for multi-speaker
attention_typeCOMBOautoAttention implementation. Auto selects the best available, eager is standard, sdpa is optimized PyTorch, flash_attention_2 requires compatible GPU, sage uses quantized attention for speedup (CUDA only)
quantize_llmCOMBOfull precisionDynamically quantize only the LLM component for non-quantized models. 4bit: major VRAM savings with minimal quality loss. 8bit: good balance of quality and memory usage. Full precision: original quality. Note: ignored for pre-quantized models. Requires CUDA GPU.
free_memory_after_generateBOOLEANtrueFree model from memory after generation to save VRAM/RAM. Disable to keep model loaded for faster subsequent generations
diffusion_stepsINT201–100Number of denoising steps. More steps = theoretically better quality but slower. Default: 20
seedINT420–4294967295Random seed for generation. Default 42 is used in official examples
cfg_scaleFLOAT1.300.5–3.5Classifier-free guidance scale (official default: 1.3)
use_samplingBOOLEANfalseEnable sampling mode. When False (default), uses deterministic generation like official examples
speaker1_voiceoptAUDIOOptional: Voice sample for Speaker 1. If not provided, synthetic voice will be used.
speaker2_voiceoptAUDIOOptional: Voice sample for Speaker 2. If not provided, synthetic voice will be used.
speaker3_voiceoptAUDIOOptional: Voice sample for Speaker 3. If not provided, synthetic voice will be used.
speaker4_voiceoptAUDIOOptional: Voice sample for Speaker 4. If not provided, synthetic voice will be used.
loraoptLORA_CONFIGOptional: LoRA configuration from VibeVoice LoRA node
temperatureoptFLOAT0.950.1–2Only used when sampling is enabled
top_poptFLOAT0.950.1–1Only used when sampling is enabled
voice_speed_factoroptFLOAT1.000.8–1.21.0 = normal speed, <1.0 = slower speed, >1.0 = faster speed (applies to all speakers)

Outputs (1)

NameTypeDescription
audioAUDIO