Does Google AI have an interface for recording piano sound with a microphone and converting it into midi format?

[zhang chao]

Does Google AI have an interface for real-time recording of piano sound through a microphone and converting it into midi format

[Joseph Weidinger]

AFAIK they haven't made any interface to convert a stream of audio to midi. I did an experiment a while ago buffering live audio then making requests to a local running instance of the model before finally stitching it all together. You'd probably have to make the buffers overlap quite a bit to make sure the onsets are in each buffer. It's hacky for sure but it can work depending on what your bar for quality is.

Maybe others have better ideas.

[zhang chao]

Thank you for your reply. I plan to break down the recorded audio into small segments for escape. It is obvious that there are bugs, and there is a possibility of data loss due to improper segmentation.