facebookresearch/seamless_communication ? reverse-engineered prompt

Build me a Python project that lets me translate speech and text across languages using Meta’s Seamless models.

I want a simple command line experience and a small demo app so I can give it either an audio file or plain text, choose a source and target language, and get back translated speech or translated text. It should support speech to speech, speech to text, text to speech, text to text, and basic speech recognition. If possible, include the expressive and streaming versions too, so I can test real time translation and more natural sounding voice output.

Make sure the project can be installed easily, comes with clear examples, and includes a notebook or tutorial that shows how to run the main workflows end to end. If there are any current setup details or model requirements that might have changed, look them up online and use the latest docs.