an annoying amount of modern infowork is some sort of transcription++...eg, paper notes into cleaned up digital notes, or personal notes into notes for sharing, or photos of whiteboards into notes... they look like they should be AI-automatable like audio transcription but aren't
Conversation
Replying to
often the reason to transcribe is to put them into a reviewable/resurfaceable format, and/or link them up to other notes
1
12
This Tweet is from an account that no longer exists. Learn more
Show replies
Replying to
Shape detection by color separation & hard edges. Those ML techniques separate->clone entities into vector/rasterized lines & solid color shapes (content-aware fill where text was)
You'd have separated, draggable objects. Then use another ML model to turn text images into text
Replying to
Cheap realtime speech transcription has gotten decent, but converting audio files to text files for cheap/free is still problematic. There are some workarounds here:


