//
sign in
Post
by @danabra.mov
PostEmbed
by @danabra.mov
Record
by @jimpick.com
Record
by @atsui.org
+ new component
Post
I’ve been playing with tiny ocr and transcription models lately that run in wasm (or in Claude’s skill compute layer) They’re not 100% accurate, but they’re USEFUL, especially when output is fed to Claude Tried transformer versions of both; useLESS! Makes me wonder what else I’m missing out on?