Update main.py for Kokoro

Revert to working Piper version

Restore Kokoro files

Test minimal build

Replace Piper with Kokoro 82M TTS - higher quality CPU TTS

Fix Piper 1.2+ synthesize API: handle single-speaker models (no sid input)

Fix Piper 1.2+ synthesize API: use wave.open with BytesIO buffer

Fix Piper 1.2+ synthesize API: pass wav_file BytesIO buffer

Piper TTS: runtime model download, no build-time network

Switch to Piper TTS - lightweight, fast, multilingual

Replace RVC WebUI with Matcha TTS API - faster TTS solution