Choose Cloud or Local models

Choose where Dictation, Actions, Notetaker, meeting summaries, and Narrator process your audio and text.

Each Epilude feature has its own model setting. Open Settings → Models to choose a Cloud model or Local model for every available feature.

Choose a model for each feature

Changing one model does not change the others. For example, you can dictate with a Local model, transcribe meetings in the cloud, and listen with a voice that runs on your Mac.

FeatureCloud modelLocal model
DictationAvailableAvailable on Apple Silicon Macs
ActionsAvailableNot available yet
Notetaker transcriptionAvailableAvailable on Apple Silicon Macs
Meeting summariesAvailableAvailable on 16GB+ Apple Silicon Macs
NarratorPremium voicesEP-1 voice (download) or macOS system voices

If Local is unavailable, the selector remains visible but disabled. Hover over the information icon beside a selector to compare its models and requirements.

Set the Dictation model

The Dictation setting controls transcription and Cleanup while you dictate. Cloud uses Epilude's cloud model, while Local runs dictation on your Mac with EP-1.

Cloud dictation offers real-time transcription and the widest language support. Epilude sends your audio for transcription but never saves your dictations on its servers.

Local dictation keeps your voice and text on your Mac and works offline after installation. It requires an Apple Silicon (M-series) Mac. Full on-device Cleanup also requires at least 16GB of memory; Macs with less memory use on-device transcription and basic formatting.

Download the Local model

Open Models settings

Open Settings → Models, then find Dictation.

Select Local

Select Local beside Dictation model. The On-device models section appears below the selector.

Download the model

Select Download beside EP-1. Dictation is paused in Local until the download finishes, but you can switch back to Cloud while it downloads.

The downloaded model is shared with local Notetaker transcription. Removing it pauses both features whenever they are set to Local.

Set the Actions model

Actions uses the Cloud model. When you run an Action, Epilude sends the text you selected and your spoken command to its cloud model.

The Local option is disabled because Actions does not have an on-device model yet. This setting stays independent from the Dictation setting, so choosing Local for Dictation does not change Actions.

Set the Notetaker models

Notetaker separates transcription from meeting summaries. You can choose Cloud or Local for Transcription model and, on a 16GB+ Apple Silicon Mac, for Meeting summary model too. The two choices are independent.

With Local transcription, meeting audio and the transcript stay on your Mac. Cloud transcription sends meeting audio to Epilude for transcription and requires an internet connection.

With the cloud summary model (the default), stopping a recording sends the finished transcript to Epilude's cloud model to produce notes and action items; this step does not send meeting audio. With Local, EP-1 writes the notes on your Mac and nothing about the meeting leaves it. Choosing Local shows what you give up first: about 6 in 10 on-device notes come out ready to send, against about 9 in 10 with the cloud model, and a 45 to 60 minute meeting takes about 20 minutes to write up. It also starts the one-time download of the summariser part of EP-1 (about 4.3 GB). Every note says which model wrote it.

Set the Narrator model

Narrator offers premium Cloud voices and two kinds of voice that run on your Mac. Cloud narration sends the text to Epilude and requires an internet connection.

With Local, Narrator reads with the EP-1 voice on Apple Silicon Macs, or with the macOS system voices. The EP-1 voice is a one-time download of about 100 MB for English, with a library of further English voices as a separate small download; German, Spanish, French, Italian and Portuguese download separately when you pick them in Settings → Narrator. Text in any other language is read with a system voice, and the player says so. Either way, the narration text stays on your Mac and works offline. Putting a narration on the web is the one thing that uploads, and only when you ask for it: by sharing it, or by choosing Keep on the Web. The setting that keeps narrations automatically applies to cloud voices only, never to these.

Check what leaves your Mac

Each feature has a separate data boundary based on its selected model.

Feature and modelSent to EpiludeStays on your Mac
Dictation, CloudDictation audio for transcriptionAudio is not stored on Epilude's servers
Dictation, LocalNothingDictation audio and text
Actions, CloudSelected text and spoken commandOther content in the current app
Notetaker transcription, CloudMeeting audio for transcriptionThe saved recording in Epilude
Notetaker transcription, LocalNothingMeeting audio and transcript
Meeting summaries, CloudFinished transcriptMeeting audio
Meeting summaries, LocalNothingMeeting audio, transcript, and notes
Narrator, CloudNarration textOther text on your Mac
Narrator, LocalNothing, unless you put a narration on the web: then its text and audio, when you ask for itNarration text and generated speech

Troubleshooting

  • Local is disabled. Local Dictation and Notetaker transcription require an Apple Silicon Mac, and on-device meeting summaries need a 16GB or larger Apple Silicon Mac. Actions has no Local model.
  • Local Dictation or Notetaker is paused. Open Settings → Models → Dictation, then download EP-1 under On-device models.
  • Cloud is unavailable. Cloud features require an active Pro entitlement after the trial or a plan that includes cloud processing.
  • A language does not work locally. Switch that feature to Cloud for languages outside the Local model's supported European languages.

Next steps

Was this page helpful?