k2-fsa/sherpa-onnx ? reverse-engineered prompt

Build me a simple offline speech app using sherpa onnx that can run on my computer without needing the internet.

I want the app to let me record or upload audio, then show live speech to text as the person talks. It should also be able to detect when someone starts and stops speaking, and if possible show speaker changes in a meeting style recording. Add a text to speech panel too, so I can type a message and hear it spoken out loud.

Please make it easy to run on desktop first, with a clean interface and clear buttons for start, stop, transcribe, and speak. If you need to, look up the current docs online and use the best model setup for local use. Also include a short readme that explains how to build and run it on my machine, plus a few example commands or steps for testing with sample audio.