Automate end-to-end voice cloning workflows on Apple Silicon with Qwen-tts.
npx skills add https://github.com/hewliyang/qwen-tts --skill qwen-tts-voice-cloning
Please help me install this Agent Skill. Skill: qwen-tts-voice-cloning Source: https://github.com/hewliyang/qwen-tts/tree/main Command: npx skills add https://github.com/hewliyang/qwen-tts --skill qwen-tts-voice-cloning
This Skill provides an end-to-end workflow to clone a voice using Qwen-TTS on Apple Silicon, encapsulating data collection, cleaning, preparation, training, generation, and evaluation in a reproducible pipeline.
Begin by collecting audio, splitting, preparing data, training a speaker, generating speech, and evaluating results with the CLI.
High-intent search queries and answers about installing and using this skill.
End-to-end voice cloning on Apple Silicon is automated through a pipeline covering data collection, cleaning, preparation, LoRA fine-tuning, generation, and evaluation using the Qwen-tts toolchain.
Audio data must be in WAV format at 24kHz mono to train custom voice models with Qwen-tts, ensuring proper preprocessing for ASR-based transcription and LoRA fine-tuning.
Single-speaker voice cloning supports base, voice-design, and custom-voice model variants, available in 0.6B and 1.7B parameter sizes for flexible training and generation.
Voice cloning evaluation applies local ASR intelligibility checks using parakeet-tdt alongside Gemini-based speaker similarity checks, providing automated quality assurance for generated speech.
Before starting the voice cloning pipeline, you need parakeet-tdt ASR, the qwen-tts toolchain, and Gemini for evaluation, plus audio data in WAV format at 24kHz mono.
lidge-junDownload YouTube videos and audio using yt-dlp with configurable formats.
refly-aiGenerate speech, transcribe audio, and create voice narration with Fish Audio models.
tdiminoTranscribe audio files offline using NVIDIA Parakeet TDT models.
EmZodGenerate speech from text with on-device TTS and voice cloning on Apple Silicon.
