Energy and battery life
When YapToText is idle it is not doing much: CPU use returns to about zero between dictations, and after a delay you choose the models are released from memory. What it costs the rest of the time comes down to three settings: which models run on battery, how long the models stay loaded after a dictation, and whether the microphone stays warm. This page covers all three and what each one trades away.
Everything runs on the Mac itself, so the only question is how much of your own Mac the app may use, and when.
What the Energy page shows
The Energy page in Settings starts with what this Mac is: the chip, the memory, the core count, the model identifier, and whether it is plugged in or on battery right now. On a laptop the power line also shows the battery percentage. A desktop Mac has no battery, so it always reads as plugged in, and the power-source features below simply stay on the plugged-in side.
Models that follow the power source
Turn on Switch models with the power source and the app runs different models depending on how the Mac is powered. The dictation model, the one that turns your voice into words, and the cleanup model, the one that turns those words into finished text, are set separately. Each has a plugged-in choice and an on-battery choice. Every model you have shows up in those menus, including ones you added yourself.
Leaving a menu on Use the main selection keeps whatever is chosen on the AI Models page, so you can change only the battery side and leave the rest alone. A mode that carries its own model choice in its editor always wins over these defaults.
The cleanup model is the heaviest thing the app runs, so putting a lighter one on battery saves the most power. The AI Models page shows the size of every model in its two libraries and rates each one for accuracy and for performance. Bigger models hear better; smaller ones are lighter on battery and memory.
The switch is off by default.
Model cooldown: how long models stay in memory
A loaded model answers instantly but holds memory the whole time it sits there. The Keep models in memory setting, under Energy on the Advanced page, decides how long the speech and cleanup models stay loaded after you finish a dictation. The choices are Only while dictating, 10 seconds after use, 30 seconds after use, 2 minutes after use, 15 minutes after use, and Until quit. The default is 2 minutes.
Unloading sooner saves energy and memory. The cost is that the next dictation after an unload takes a few extra seconds while the model reloads, and the panel tells you while that happens. Only while dictating is the deepest saver. Until quit is the fastest, and it holds the memory for as long as the app is open.
The models also warm up at launch, so the first dictation after launch does not pay the whole load, though it can still take a moment. Whenever a model is loading, the panel tells you. The Statistics page has an Energy & resources readout: memory right now, CPU right now, and the model unload delay. Memory is mostly the loaded models and drops once they are released. CPU spikes briefly during transcription and cleanup, then returns to about zero at idle.
Keep the microphone warm
Keep the microphone warm lives in the Microphone section of Preferences, and it is off by default. Off means the microphone lets go about a second after each dictation ends, and the orange indicator only shows while you dictate. The app's own note on that setting says the next dictation may then miss its first second.
On means the input stays open after a dictation so the next one starts instantly. A Stay warm for slider sets how long, from 1 to 30 minutes, starting at 5. When that window runs out the microphone genuinely lets go, the same as if the setting were off. While it is warm the mic indicator stays on, which is the honest cost: an open input and a light that stays lit. Warm audio is analyzed for levels only and is never stored.
Launching or switching apps never opens the microphone on its own.
What each one costs
- A heavier model on battery. Better hearing and better cleanup, paid for in battery and memory during every dictation. The cleanup model is the single biggest cost.
- Keeping models in memory longer. Instant answers, paid for in memory held while you are not dictating. A shorter delay gives the memory back sooner and costs a few seconds of reload on the next dictation.
- A warm microphone. An instant start on every dictation, paid for with an open input and a lit indicator for the minutes you chose.
- A dictation itself. A brief CPU spike for transcription and cleanup, then back to idle.
If I wanted the longest battery life I would put a lighter cleanup model on the battery side, set Keep models in memory to Only while dictating, and leave the microphone cold. If I wanted the fastest possible dictation I would do the opposite: Until quit and a warm microphone. The defaults sit in between: the same models on both power sources, a 2 minute cooldown, and a cold microphone.
Questions
Does YapToText use battery when I am not dictating?
Very little. CPU returns to about zero at idle, and once the Keep models in memory delay passes the models are released and the memory drops. A warm microphone is the one setting that keeps the input open between dictations, and it is off by default.
Why is my first dictation after a break slower?
The models were unloaded after the cooldown and have to reload, which takes a few extra seconds. Choose a longer Keep models in memory delay on the Advanced page, or Until quit, if you would rather hold the memory and skip the wait.
Why does the orange microphone indicator stay on after I stop?
Keep the microphone warm is on, so the input stays open for the minutes set on the Stay warm for slider and the next dictation starts instantly. Turn it off in the Microphone section of Preferences and the microphone lets go about a second after each dictation.
Can I run a smaller model only on battery?
Yes. Turn on Switch models with the power source on the Energy page and pick a lighter model in the On battery menu for the dictation model, the cleanup model, or both. Leave the Plugged in menu on Use the main selection to keep your usual choice there.
Does the Energy page do anything on a desktop Mac?
It still shows the chip, memory, cores, and model identifier, but a Mac with no battery always reads as plugged in, so the on-battery model choices never apply. The model cooldown and warm microphone settings work the same on any Mac.