KoljaB/RealtimeSTT ? reverse-engineered prompt

I want to build a Python script that listens to my microphone and turns my speech into text in real-time. It should feel instant, with very low latency, so I can see the words appear right after I say them. The script should be smart enough to detect when I start and stop talking, so it isn't constantly processing background noise.

I'd also like a "wake word" feature, so it only starts transcribing after I say a specific command. A really cool use case would be for it to type whatever I say into any active window on my computer, like a universal dictation tool. For the transcription, please use a fast version of the Whisper model. It's important that this is optimized for performance, so make sure it's set up to use my NVIDIA GPU with CUDA. You can look up the latest docs for setting that up if you need to.