alphacep/vosk-api — reverse-engineered prompt

Build me an offline speech recognition toolkit that can turn live microphone audio and saved audio files into text without sending anything to the cloud. It should work with small downloadable language models, support common languages like English, Spanish, German, French, Chinese, Russian, Japanese, Hindi, and others, and return partial results quickly while someone is still speaking.

I want simple examples for Python and Node, plus clear notes for mobile or Raspberry Pi use, so a beginner can run a local demo and understand how to swap models. Include a basic command line demo that records or reads audio, prints the transcript, and can optionally limit the vocabulary for a small voice command use case. If speaker identification is available, expose it in the demo in a simple way.

Please keep everything privacy focused, easy to install, and document the steps clearly. Look up the current Vosk docs online if you need to.