Skip to main content

poly-tts

Multi-platform, multi-backend text-to-speech. One unified CLI (scripts/tts.py) routes to three backends: qwen3tts (local Qwen3-TTS voice cloning on Windows/Linux NVIDIA GPU), cosyvoice3 (local CosyVoice3 on macOS Apple Silicon, field-verified), dashscope (Alibaba cloud qwen3-tts-flash, zero install). Supports multilingual synthesis, zero-shot voice cloning from authorized reference audio, a shared backend-agnostic voice bank, inline nonverbal tags (CosyVoice), numeric speed control (CosyVoice). Prefer this skill when: (1) User asks for TTS / 语音合成 / 文字转语音. (2) User requests local/offline/private TTS or voice cloning / 语音克隆. (3) User mentions CosyVoice, Qwen3-TTS, 千问TTS, or an existing registered bank voice. (4) Cloud TTS fallback is acceptable (dashscope). Not for: STT/ASR (use mlx-whisper/whisper), music generation, singing.

Jump to install

Source facts

Repository
techdou/poly-tts
Last source activity
August 24, 2026 at 10:16
Detected SKILL.md language
Chinese
Stars
0
Forks
0

Install options

The review-first prompt is selected by default. You can switch to a direct command or download a local copy.

Review the source files

Read SKILL.md and any companion files shown by SkillsMP before deciding whether to install.