BREAKING
Qwen3-TTS Comes to ComfyUI
Audio Pipeline From Nodes
1
VoiceDesignNode
↓
2
VoiceCloneNode
↓
3
CustomVoiceNode
↓
4
DialogueInferenceNode
0
B
fast model
0
B
high quality
0
languages
Local, Flexible, Fiddly
Strengths
●
Fully local, Mac MPS supported
●
Zero-shot voice cloning
●
Multi-speaker dialogue
Caveats
●
Pin transformers==4.57.3
●
Quality depends on reference audio
●
VRAM management matters
5-15s Reference Recommended
Open-Source TTS Grows
AI NEWS BLITZ
New custom nodes bring the open-source Qwen3-TTS model into ComfyUI workflows.