qwen3-audio_skill

Overview

This skill is a high-performance audio library optimized for Apple Silicon that provides fast text-to-speech (TTS) and speech-to-text (STT). It supports multiple models, voice cloning, voice design, emotion/style control, and common audio formats for production workflows on M1–M4 Macs.

How this skill works

The library runs locally on Apple Silicon and exposes command-line scripts to perform TTS, STT, voice creation, and voice management. TTS can use reference audio or predefined speakers and supports instruction-based style control; STT produces transcripts and subtitle formats. Voices are stored as reusable profiles with reference audio and transcript files for quick cloning and reuse.

When to use it

• Generate high-quality speech locally on Apple Silicon without remote APIs.
• Clone a specific voice from a short reference recording for personalized TTS.
• Produce subtitles or transcripts from audio files in txt, srt, or ass formats.
• Create custom voice timbres by describing desired characteristics for voice design.
• Manage and reuse multiple voice profiles across projects and pipelines.

Best practices

• Run environment checks on an Apple Silicon Mac with Python 3.10+ before use.
• Use clear, noise-free reference audio and an accurate transcript for best cloning results.
• Choose a predefined speaker whose native language matches the target language for higher quality.
• Store voice profiles in the provided voices/ directory and use consistent IDs for automation.
• Export STT outputs to subtitle formats (srt/ass) when integrating with video workflows.

Example use cases

• Produce narration for short-form videos with emotion control (e.g., 'Very happy and excited').
• Create a reusable brand voice by recording a reference clip and creating a voice profile.
• Transcribe meeting recordings to editable text and srt subtitles for distribution.
• Rapidly prototype multilingual voice interactions using predefined speakers for each language.
• Convert written content to audio assets for podcasts or accessibility on Apple Silicon devices.