Just found out that the Nextcloud whisper stt model doesn't actually work with the smart picker, meaning I've been using the API this whole time 🫠

And all seriousness though it really isn't that expensive at like 0.006$ per minute. And it is really good accuracy. Just frustrated that I wasn't running it locally. 

 One day, I want to build a server with a very nice GPU that I can use for transcoding and running AI models