qwen3-tts-apple-silicon and qwen3-tts-mac
These are ecosystem siblings where the "apple-silicon" variant is a feature-enhanced fork of the "mac" implementation, both targeting the same MLX-based Qwen3-TTS inference stack on Apple Silicon but with the former adding voice cloning and voice design capabilities.
About qwen3-tts-apple-silicon
kapi2800/qwen3-tts-apple-silicon
Run Qwen3-TTS text-to-speech locally on Mac (M1/M2/M3/M4). Voice cloning, voice design, custom voices. 100% offline using MLX.
Built on MLX's Apple Neural Engine integration, the implementation uses 8-bit quantized Qwen3 models (1.7B for quality or 0.6B for speed) that reduce RAM overhead to 2-3GB while maintaining native GPU acceleration. The CLI interface provides three distinct inference pathways: preset voice synthesis with emotion/speed modulation, text-conditional voice generation, and speaker embedding extraction from reference audio for cloning.
About qwen3-tts-mac
kapi2800/qwen3-tts-mac
Optimized implementation of Qwen3-TTS for Apple Silicon (M1-M4)
Related comparisons
Scores updated daily from GitHub, PyPI, and npm data. How scores work