Feedback: this tool is just great

[jerkstorecaller]

I'm getting MUCH nicer and more accurate subs with 'tiny' on stable-ts than I do with 'medium-q8' on whisper.cpp and pywhispercpp, in a TV show with multiple speakers.

I know you guys are all using the same underlying tech, I know I can probably get close to stable-ts by learning the ins and outs of those other tools, but stable-ts "Just Works" with whatever defaults and QoL features it does under the hood. And I didn't have to learn anything (except see next paragraph). With some better examples it's something we can point the average non-techy geek at to create subs of any audio they can't find subs for. I intend to contribute this doc once I have some time off and get to use it more.

The one thing I found annoying and confusing is having karaoke mode enabled in the CLI when saving subtitles. I don't mean the --karaoke param (which is confusingly named considering), but word_level defaulting to True when it should be False. Think about the number of people who want subs for tv/movies/podcasts and then the number who want karaoke/learning languages. I bet the latter are less than 0.1%. Anyway, this is one of the newbie traps I will cover in a guide.