🎙️ TTS-Skill — Multi-Engine Text-to-Speech

TTS-Skill is a unified entrypoint that wraps multiple TTS engines:

• Qwen3-TTS: local voice cloning (reference audio + reference transcript)
• VoiceCraft / Edge TTS: online Microsoft-style voices with speed/pitch/style controls
• OpenAI TTS: commercial-quality speech generation via API

Key Features

• Multi-engine routing: choose the best engine per use case
• Voice matching: pick a voice by filename keyword (for local cloning)
• Progress feedback: long-running Qwen3-TTS jobs show a live progress bar and ETA
• Consistent outputs: default output goes to output/ with a predictable name

Default Output

If --output is not provided:

• Output directory: output/
• Filename: YYYYMMDD_HHMMSS_<first-6-chars>.<ext>

Quick Start (CLI)

# Local voice cloning (Qwen3-TTS)
python tts-skill.py qwen3-tts "胜利在呼唤，勇往直前" --voice 赵信

# Online voices (VoiceCraft / Edge)
python tts-skill.py edge-tts "你好，这是一个测试" --voice xiaoxiao

# OpenAI TTS
python tts-skill.py openai-tts "Hello, this is a test" --voice alloy

Read long text from a file:

python tts-skill.py qwen3-tts --text-file "input\\text.txt" --voice 寒冰射手

Voices

Local (Qwen3-TTS)

Place a pair of files in assets/:

• assets/<VoiceName>.wav (or .mp3/.m4a/.flac)
• assets/<VoiceName>.txt (the transcript matching the audio)

Then use:

python tts-skill.py qwen3-tts "测试文本" --voice <VoiceName>

VoiceCraft / Edge

python engines/edge-tts-cli.py --list-voices
python engines/edge-tts-cli.py --list-styles

OpenAI

Create engines/openai-tts.config and set api_key.

Docs

• Installation: INSTALL.md
• Skill spec: SKILL.md
• Chinese README: README.zh-CN.md

Acknowledgements

This project uses components from https://github.com/wangwangit/tts. Special thanks to the original authors for their contributions to the TTS community.