Arka-10717/ComfyUI-Qwen-TTS
🗣️ Create custom nodes for speech synthesis, voice cloning, and voice design using the open-source Qwen3-TTS framework.
What's novel
🗣️ Create custom nodes for speech synthesis, voice cloning, and voice design using the open-source Qwen3-TTS framework.
Code Analysis
8 files read · 3 roundsA ComfyUI custom node package that wraps Alibaba's Qwen3-TTS model (including ECAPA-TDNN speaker encoder, VQ speech tokenizer, and transformer-based TTS generation) to provide text-to-speech, voice cloning, and voice design capabilities within the ComfyUI node-based workflow environment.
Strengths
The code is well-structured with clear separation between model definitions, inference wrappers, and ComfyUI node integration. The inference layer provides a clean HuggingFace-style API with proper type hints, comprehensive docstrings, and robust input normalization supporting multiple audio formats (paths, URLs, base64, numpy arrays).
Weaknesses
The README is catastrophically misleading — it's AI-generated filler with broken links and nonsensical instructions that bear no resemblance to the actual ComfyUI node-based workflow. There are zero tests despite the project having 48 source files and complex ML pipeline logic. The nodes.py file is a 59kb monolith that mixes model management, UI definitions, and monkey-patching.
Score Breakdown
Signal breakdown
Innovation
Craft
Traction
Scope
Evidence
Commits
68
Contributors
7
Files
43
Active weeks
4
Repository
Language
Python
Stars
1
Forks
0
License
—