superwhisper is slow or uses too much memory: the settings that change it

By · Published

Slow dictation in superwhisper comes from one of two steps: turning your voice into text, or the AI processing afterward. Each step has its own model, and each can be local or in the cloud. superwhisper's docs give a way to find which step is slow, and a few settings that fix most cases.

I built YapToText, a dictation app for the Mac, so read what I say about superwhisper with that in mind. Every fact about superwhisper here comes from its own docs, read on October 11, 2026.

Download YapToText on the Mac App StoreThis app is completely free. You don't have to make an account. There are no subscriptions, no difficult downloads, and nothing is locked. Made for Apple silicon Macs. Source on GitHub.

Find the slow step first

Open a dictation in History and compare the voice and AI processing times in the right sidebar. Whichever step is slow, switch that model. The rest of this page is the list of what to switch to.

Local models: memory and loading

  • Voice model active duration. In Settings, Advanced settings, under Voice model. It sets how long a voice model stays loaded, from 10 seconds to 1 hour. Shorter frees memory but reloads the model each time. Longer keeps it ready and holds memory.
  • Free memory. Local models need it. Close memory-heavy apps while dictating.
  • A smaller model. If the Mac still struggles, the docs suggest Whisper Base, or switching that mode to a cloud model.

Cloud models: your connection

Cloud models do the work on remote servers, so your hardware matters less and your connection matters more. On a slow or unstable connection, the docs say a local model can work better.

Three ways to speed it up

  1. Skip AI processing when you do not need it. The Voice to Text mode runs only the transcription step.
  2. Try a different voice model. The docs name S1-Voice as a cloud option that moves the work off your machine, Parakeet Multilanguage, at 494 MB, as a local one that runs offline, and Whisper Tiny or Whisper Base as smaller local ones.
  3. Check History again and see whether the slow step moved.

The best superwhisper model page compares the models in more detail.

If it crashes instead of slowing down

The docs treat crashes with a local model as a memory problem too: switch that mode to a cloud voice model, or a smaller one such as Whisper Base, and see whether the crash stops. A nearly full disk also causes failures. Your dictation is usually safe, because superwhisper writes audio to disk as you record and a crashed dictation tends to appear in History. superwhisper not working covers the rest.

Speed in YapToText

YapToText has the same two steps, speech then cleanup, both on the Mac. Its History shows the stop-to-text time split into transcription, AI cleanup and delivery for each dictation, so you can see the slow step the same way. Raw Transcription skips the cleanup step entirely, a mode can use a lighter speech model on battery, and Keep the microphone warm makes the next start instant. It has no cloud option to fall back on, so on an older Mac a smaller model is the only lever. The models help page lists them.

Download YapToText on the Mac App StoreThis app is completely free. You don't have to make an account. There are no subscriptions, no difficult downloads, and nothing is locked. Made for Apple silicon Macs. Source on GitHub.

The short version

  • Diagnose. History, right sidebar, voice against AI time.
  • Memory. Shorten Voice model active duration, or use a smaller model.
  • Speed. Voice to Text mode skips AI processing.
  • Old Mac. A cloud voice model moves the work away.
  • Bad connection. A local model.

If superwhisper changes these settings, open an issue on GitHub and I will update this page. Every one gets read.

Download YapToText on the Mac App StoreThis app is completely free. You don't have to make an account. There are no subscriptions, no difficult downloads, and nothing is locked. Made for Apple silicon Macs. Source on GitHub.

About the author. Ryleigh Newman is a developer whose hands make typing difficult and who dictates every day with YapToText, the app this site describes. Facts about other apps come from their own sites, with the date they were read. More about the project.

Sources

Questions

Why is superwhisper slow?

Either the voice model or the AI processing step is slow. Open a dictation in History and compare the two times in the right sidebar, then change that model.

How do I reduce superwhisper's memory use?

Shorten Voice model active duration in Settings, Advanced settings, or pick a smaller local model such as Whisper Base.

What is Voice model active duration in superwhisper?

How long a voice model stays loaded after use, from 10 seconds to 1 hour. Shorter saves memory, longer starts faster.

What is the fastest superwhisper setup?

Its docs suggest the Voice to Text mode, which skips AI processing, with a fast model such as Parakeet Multilanguage locally or a cloud voice model.

Does YapToText show where the time goes?

Yes. History splits each dictation's time into transcription, AI cleanup and delivery.