parrots

Why we included this project

Parrots bundles speech recognition and text-to-speech into one Python package, so you don't have to wire separate ASR and TTS models together yourself. Recognition is handled by a distil-whisper-based model covering Chinese and English, with pinyin-to-character support, while synthesis offers three routes: a GPT-SoVITS-trained multi-voice engine, the IndexTTS2 model with decoupled emotion and speaker control plus adjustable duration, and streaming output for low-latency real-time speech. That makes it attractive if you're building Chinese- or English-language voice features and want both directions covered by a single library instead of assembling your own pipeline. Multi-lingual TTS with fine-grained control over emotion and speaking rate is the strongest selling point, and the streaming path gives you a fast path for real-time output. It installs with pip and has a Hugging Face demo if you want to hear the quality before committing.