One step closer to figuring out live autocaptions that might be semi-tweakable! I adapted some code from https://github.com/deepgram/streaming-test-suite/ to let me also automatically save the JSON and text for further processing. I think I’ll be able to use start-process in Emacs to get that to listen to my audio and put the text in a buffer, so we can get live notes during streaming or braindumping. If I can use Alsa to pipe audio into the process, I might be able to rig it up to send lines to an IRC channel using ERC or overwrite a text overlay that OBS uses, so a future EmacsConf might even have auto captions for live talks. Bonus points if I can someday figure out how to correct misrecognized words on the fly, either by pattern-matching on common errors or having a quick way I can replace a word or two…
sachac's Blog
Interests include: #Emacs, #OrgMode, #elisp, #nodejs, #python, #sketchnotes, #parenting, #cooking, #gardening, #knitting, #sewing, #lego, #captioning, #plover #steno, and #stoic philosophy. Originally from Manila, now in Toronto. Married to a Vim guy (go figure) and raising a 7-year old (editor preference unknown), along with two very loud cats.
Blog: https://sachachua.com (mostly Emacs News these days), sketches: https://sketches.sachachua.com. I also maintain planet.emacslife.com and subed.el
Blog: https://sachachua.com (mostly Emacs News these days), sketches: https://sketches.sachachua.com. I also maintain planet.emacslife.com and subed.el