Product Suggestion: User experience when recompiling model
A minor but frequent pain point is that sometimes I am in the middle of working, in the middle of a thought, I am jamming, dictating, copying and pasting, editing. Then I get hit with a model recompile, and the whole world has to stop and wait. I didn't initiate a software update. I was just using the software and then it became unavailable for reasons that are not unavoidable.
I think it would be better to enqueue a model recompile, and then do it AFTER you finish the current transcription job. So instead of capturing my audio, then showing me "recompiling model, please wait..." the system transcribes my result normally and immediately, and then starts the recompile AFTER I am done with my transcription task. I can then swipe or click away, and the software probably has enough time to finish recompiling in the background before it is paged out, put to sleep, etc. If you make model compilation progressive, you don't even have to finish it in one job. You can just have a background thread work on it in increments until it finishes with the time it has.
This would take model recompilation from a periodic nuisance to something you could brag about to the user. Instead of a modal "WAIT WHILE WE RECOMPILE" you could show periodic "btw, I finished compiling the latest model for you, and it's ready to go! 🏎️" which is a much more positive interaction.
This would make using Superwhisper a lot smoother for me. I'd appreciate something like this, or some change that makes model compilation less "my problem" and more something the software handles for me seamlessly.
Thanks for your time!
Pending