neuphonic/neutts — reverse-engineered prompt

Build me a simple Python text to speech project that runs locally on someone’s device, not through a cloud voice API. I want users to type text, pick one of the supported NeuTTS style models, and generate natural sounding speech as a wav file.

It should support the main use cases from the NeuTTS README, including English and the Nano models for Spanish, German, and French when those models are selected. Please include instant voice cloning where the user can provide a short voice sample, around 3 seconds, and then generate speech in that voice.

Make it easy to run from a basic example script, and include a streaming example too if possible. It should work with Hugging Face model downloads and the smaller GGUF options for local inference. Please keep setup straightforward, with clear install instructions for normal Python users and notes for Mac, Linux, and Windows if extra local performance setup is needed.

Look up current docs online if you need to.