Really hope I’m wrong about this, but this app for speech-to-text transcription appears to be constantly using 200% - 430% CPU to do nothing other than write progress to a SQLite database and update its UI. The transcription is happening in Apple’s localspeechrecognition process.

I took a few samples, and it seems to be receiving a stream of progress updates from Speech.framework, then writing json of each transcript line to SQLite, then SwiftUI fires a re-render, where it goes and queries thousands of rows from SQLite and decodes all that json. Happens even if no transcript is being shown. What a mess.

This is an app with tons of features, plenty of models supported, etc, which sells at $29.99/year or $99.99 for lifetime (USD). Do they even realise their app is doing this?

If anyone has suggestions for decent apps that can batch transcribe with Apple’s model and export to .srt/.vtt, would really appreciate it.