OpenAI-compatible TTS endpoint using OmniVoice: 0.3s response time

I want to share a TTS server with an OpenAI-compatible API that generates speech really fast (about 0.3 seconds for a sentence on an RTX 3080) and can clone a voice from a short reference clip. I’ve optimized the server so generation starts quickly, and it processes long text in sequence, para

中文全文翻译正在补齐，可先查看下方英文原文。

I want to share a TTS server with an OpenAI-compatible API that generates speech really fast (about 0.3 seconds for a sentence on an RTX 3080) and can clone a voice from a short reference clip. I’ve optimized the server so generation starts quickly, and it processes long text in sequence, para