iamdinhthuan/Kokoro-Vietnamese ? reverse-engineered prompt

Build me a Vietnamese text to speech project based on Kokoro that can take Vietnamese text and turn it into natural sounding speech with a few different voices. I want a simple Python API for generating audio and getting the phonemes back, plus a command line tool that can synthesize one sentence or a whole batch of text files into WAV files.

It should support downloading the needed model files automatically when I do not provide local paths, but also let me point to local model, voicepack, and config files if I already have them. Please include a way to list available voices, choose a voice, and optionally limit peak volume before saving audio. If it makes sense, add ONNX export and ONNX inference too, and a small Gradio interface for trying it in the browser.

Use current docs online if you need to, and make sure the install and training instructions are clear enough for someone to follow without much hassle.