Local and cloud processing

By default, everything happens on your computer: transcription, summaries, dictation cleanup, and search. After the models download once, none of it needs a network connection. There’s no account unless you want one.

embral cloud is the optional paid tier. It transcribes with higher accuracy and less battery use, labels up to 15 speakers live, and writes summaries with a larger model.

What leaves your machine

With cloud processing on, audio streams to embral’s servers during a recording or dictation, and the transcript comes straight back. For cloud summaries and cleanup, only the transcript goes up, and only the result comes back. The servers store neither. They meter usage in seconds and request counts, and that’s all they keep. Everything in your library lives only in your own storage folder.

Four separate choices

The cloud choice is per feature, not all-or-nothing: meeting transcription, dictation transcription, summaries, and dictation cleanup each have their own setting. Signing in with an account that has hours switches all four to cloud. Signing out switches them back. You can mix freely afterward, for example cloud meetings with on-device dictation.

Hours

Meetings and dictation spend from one shared pool of cloud hours. The ring in the sidebar shows what’s left. When the pool runs dry, the When cloud hours run out setting decides what happens: switch to your device (the default) or stop transcribing. Either way, the recording itself continues. See Turn on cloud processing.