iojh2021-oss/alad-persian-live-dub ? reverse-engineered prompt

Build me an Android app that can listen to audio playing from other apps, like YouTube, and dub it live from English into Persian. I want it to capture the other app’s playback audio, send that sound to Gemini Live, and play back the generated Persian speech in near real time while keeping the original audio a bit quieter in the background.

The app should let me enter my Gemini API key, start and stop dubbing easily, and ask for the needed screen and audio capture permissions when I tap start. Please make it work with the current Gemini Live websocket flow, use raw audio in and out, and handle reconnects so the session can continue smoothly if the connection drops. I only need something practical and simple for personal use, but please structure it cleanly so it could be adapted later. If you need to check the latest Gemini docs while building it, go ahead and do that.