Hacker News new | ask | show | jobs
by onaclov2000 11 days ago
I've not used it excessively on webpages, the ones I have, I start reading then change my mind and decide to listen, but haven't observed the emoji issue. Then again, maybe I tend to skip over listening to AI generated (or at least overtly obvious?) content. I also have dealt with listening to technical content on PDFs and it sometimes tries to read out x's and o's from a plot, particularly annoying driving ;) guess it hasn't ever bothered me enough to just not use it. I have slowly experimented here and there looking to build a pipeline of text ingest (thinking I could use LLM) to convert certain text (like a plot in my earlier example) into a verbal description, a plot about this thing with emphasis on showing this fact. Or basically hey take this text and make it more digestable by audio, but I haven't experimented a ton.

Ultimately yes garbage in garbage out, but I've found enough value in using it, it's been good enough for me to suggest it's use, and hasn't caused me to search for another variant of tts on phone.