What's normal
Idle, Amlin sits in the menu bar using very little CPU.
During a meeting, capture and live transcription use modest CPU — comparable to a video call.
Dictation loads a local speech model into memory when you dictate. That's the biggest single memory consumer.
Reduce dictation's memory footprint
Dictation Settings → Transcription → Resources:
Unload model after N min idle — lower it to free memory sooner. The model reloads automatically on your next dictation; the only cost is a moment's delay.
Keep microphone engine warm Ns — lower it to release the mic sooner after a session. The trade-off is that the very first word of a fast dictation may occasionally clip.
The microphone indicator stays lit
That's the warm-mic window above. It exists so your first syllable isn't cut off. Lower the warm time if you'd rather the indicator go out immediately.
Battery during long meetings
Live transcription streams audio continuously, which uses power. On a long call, plug in — same as you would for a video call.
If Amlin feels slow generally
Check whether an update is pending (Settings → General)
Quit and relaunch to clear a long-running session
If a large library feels slow, tell support — that's a bug worth knowing about.