Back to feed
Artificial intelligenceSingle source

Google AI Edge Foresight for Mac turns shorthand into full notes offline

Google introduced AI Edge Foresight for Mac, an app that processes meeting audio locally and expands shorthand into polished notes. It uses the 740-million-parameter EmbeddingGemma 2 model and can search local files and Google Drive.

Preferred on Google
Text size
Story videoWatch on YouTube ↗

Google introduced AI Edge Foresight for Mac, an app for taking notes during video and in-person meetings. It listens in the background and uses the user’s shorthand bullets and meeting transcript to produce polished notes.

The audio and transcript are processed entirely on the device, rather than in the cloud. The app can also work with project folders, reference materials, diagrams, calendars and Google Drive files.

EmbeddingGemma 2 runs locally

The features are powered by EmbeddingGemma 2, a multimodal model built on the Gemma 4 architecture with 740 million parameters. Google released it under the Apache 2.0 license and designed it for consumer hardware with tight resource constraints.

With quantization, the model uses about 191MB of active RAM for text-only weights and roughly 567MB for the full multimodal model on a Google Pixel 11 Pro.

The model can locate a specific video clip from a voice memo or search hours of audio recordings using a text query.

File search and live answers

AI Edge Foresight lets users ask questions about connected materials and retrieve information without manually organizing their files. Its chat feature supports conversational retrieval and summaries of personal knowledge.

A separate Live Assistance mode listens and automatically answers questions. EmbeddingGemma 2 demonstrations are also available in Google AI Edge Gallery for Android and iOS.

AI Edge Foresight joins Google’s other Mac apps in the series: AI Edge Gallery and AI Edge Eloquent for offline transcription.

What we know

  • AI Edge Foresight for Mac locally expands meeting shorthand into polished notes.
  • EmbeddingGemma 2 has 740 million parameters and an Apache 2.0 license.
  • On a Pixel 11 Pro, it requires about 191MB or 567MB of active RAM depending on the mode.
  • The app searches local materials and Google Drive.

Trust: Single source

  • The story relies on one publisher. A second independent confirmation has not been established from the cited sources.
  • Sources: 9to5Mac.
Related news
View sources1

COMMUNITY

Discussion

—

Loading comments…