Hello, I built this keyboard a while ago (2018), but I was sadly never able to release it. Since I had spent way too much time on this idea and I'm trying to get hired, I finally decided to atleast put it onto Github. (Usually not doing any Android development)
Thanks, yes it's exactly just that. I thought that this was missing in all those keyboards filled with features (voice recognition, translation etc).
(I have to agree that the screenshots were bad. Press the globe-button to switch into my keyboard -> get a camera-view instead of keyboard-buttons -> take a picture and tap onto a part of the text to get the input.)
I was using the "Google Mobile Vision - Text Recognition API" back then, which I think is part of their "ML Kit" now (both on-device).
The accuracy with that framework was great for full sentences/ blocks of text, probably even better with the newer version now?
(Anyone hiring for ml/mobile dev?)