I’ve been playing around with building an audiobook app, mostly through using Codex to write it, but getting in and playing with Swift code periodically as well. I wanted to clip some words out of the awesome Dungeon Crawler Carl series to make a funny “new achievement” message, and couldn’t find an easy way to find where in the ~12 hour audiobook right words happened…
… so, as you might have guessed from the title of the post, I implemented automated transcription in Bleat (more details one day soon), which took no time at all either from the code side or the actual transcription side. My iPhone 16 Pro will happily transcribe a twenty minute chapter in 51 seconds, which is more than reasonable for a battery powered device sitting in my hand!
Also, text search was two prompts - one to implement, a second to tweak the layout slightly 😊
It’s pretty amazing, just being able to use the built-in functionality of iOS to just go “hey transcribe this audio data” and the results are … as far as I can tell, perfectly usable! I haven’t tested it with multiple actors or complex words, but I’m sure that’ll be something that requries tweaking at some point if it becomes a problem.