Qwen3-TTS Voice Cloning - ComfyUI & One Click Windows Installer

The latest Qwen3-TTS Text-to-Speech AI Voice Cloning ComfyUI Workflow is now here — and it's easier than ever to get running. Featuring full Flash Attention support and an optimized one-click installer, you can start experimenting in minutes instead of hours.
Qwen3-TTS is the newest generation of text-to-speech models from Qwen AI, delivering near human-level realism in synthesized voices. Available in lightweight and large model sizes (0.6B and 1.7B), these models balance performance and quality — ideal for everything from interactive storytelling to AI voice assistants.
 
Key Highlights
Natural prosody and emotion control for dynamic speech output
Fast inference powered by FlashAttention 2.8.3 + CUDA 12.8
Flexible voice cloning with just a few seconds of audio reference
High-fidelity 12Hz generation with expressive tone shaping
 
Included in the Package
The automated installer sets up everything required to run Qwen3-TTS models through ComfyUI, including:
Flash Attention: flash_attn-2.8.3+cu128torch2.8
PyTorch: 2.8.0+cu128
ComfyUI Windows Portable
 
System Requirements
Windows OS
NVIDIA GPU with CUDA support (RTX 30XX or later recommended)
Minimum VRAM: 4 GB (more for faster processing)
Free Disk Space: At least 30 GB
FFmpeg and SOX (Sound eXchange) required
 
Usage Notes
Place the installer files in a dedicated folder, then double-click to begin setup — no extra configuration required.
Load the provided workflow inside ComfyUI.
Toggle between Preset Voices and Voice Cloning sections using the Fast Groups Bypasser above each section. Only one should be active at a time.
Preset Voices: Choose the Qwen3-TTS model (1.7B or 0.6B), enter your text, add style or tone instructions (e.g., calm and friendly, energetic podcast style), then click Generate.
Voice Cloning: Select your model, upload a 3-10 second reference clip (shorter clips are best), then click Generate.