Got it. This is happening because TalkTastic runs Whisper locally on your machine. Right now, it just sits in memory even when not in use. We've implemented dynamic model size selection based on your device to minimize memory usage - that's our v1 approach. In the next 1-2 months, we'll be unloading the model from memory when it's not being used, so it won't just sit there consuming RAM when you don't need it. Improvement is on the way!
