chidiwilliams/buzz ? reverse-engineered prompt

Build me a desktop app that can transcribe and translate audio completely offline on a personal computer.

I want to be able to open audio or video files, paste in a YouTube link, or record live from my microphone and get a transcript back fast and accurately. It should have a clean main window where I can view the transcript, search through it, play back audio while following along, and change playback speed. I also want export options for TXT, SRT, and VTT.

Please include support for different transcription backends so it can run well on different machines, including Macs and Nvidia GPUs, and make it work on Linux, Windows, and macOS. It would be great if it could separate voices or improve noisy audio before transcribing, identify different speakers, watch a folder for new files, and support a simple command line mode too. If you need to check current setup details or docs online, go ahead.