Application that records audio while a user is holding down a button, then sends that audio to some endpoint over wifi, waiting for response text back to display on the screen.
Major Components/considerations:
Audio recording
We're limited by the RP2040's low RAM, so we might have to be creative here, maybe using rp2PIO?
Likely will have to play around with bitrate/sample depth of the audio we're recording, it doesn't have to be great, just good enough for whatever's required for transcription
compression likely isn't super helpful on the badge side of things, but depending on how the backend/infra goes, it might be required. In this case, its might be worth looking into stuff re: Mel Spectrograms and mfcc (or maybe just use plain old regular compression techniques?)
Network Connection (Relies on #2 )
This might be streaming data as talking is happening, or might be just reading from a file once it's complete. Depends on the above limitations/solutions
Response Display
Probably should have the ability to scroll through the history, including transcribed audio from user (text message style?). No Images, only text.
Hackability
API Key/Endpoint/etc. should all just be in an easy config file that the use can swap out later if they want to run their own self hosted version after DC32
Depends on # and #
Application that records audio while a user is holding down a button, then sends that audio to some endpoint over wifi, waiting for response text back to display on the screen.
Major Components/considerations:
Audio recording
Network Connection (Relies on #2 )
Response Display
Hackability