Google Announces Free AI Meeting Notes App | Mac Exclusive with Fully On-Device Processing
機械翻訳 / Machine-translated

機械翻訳 / Machine-translated
@aifriends
AI Friends(https://aifriends.jp)のクロスポスト公式アカウント。AIツールの紹介・使い方・できることを、中学生でもわかるやさしい日本語で届けます。
What you'll learn in this article
On October 6, 2026, Google announced a new macOS app called "Google AI Edge Foresight" — an AI meeting notes tool that automatically transcribes meeting content and organizes notes.
Its biggest feature is that all processing happens entirely on your computer. Meeting audio and transcribed data are never sent to Google's servers. It works offline, and there's no monthly subscription fee.
It's compatible with Google Meet, Zoom, and in-person meetings. Optimized for Macs with Apple Silicon (M1, M2, M3, etc.), it is currently available as experimental software and has been released for free.
Most conventional meeting transcription tools send audio to the cloud for processing. Google AI Edge Foresight runs everything — from audio transcription to summarization — entirely on your Mac.
This makes it safe to use even in meetings involving confidential information. Whether it's a closed-door strategy session or a meeting that handles personal data, there's no risk of information leaking to the outside.
No internet connection is required. Since meeting notes can be taken offline, it works without issue even in locations with unstable network connectivity.
There are no monthly fees either. Cloud-based tools often charge thousands of yen per month to cover server maintenance costs, but Google AI Edge Foresight is free.
Google AI Edge Foresight is powered by two cutting-edge AI models: technologies called "EmbeddingGemma 2" and "Gemma 4."
EmbeddingGemma 2 is a new model announced by Google on October 6, 2026. With 740 million parameters (a figure that indicates how capable the AI is), it can simultaneously handle five types of data: text, code, images, video, and audio.
It can collectively search and analyze not just audio from meetings, but also materials shared on screen and PDF files uploaded after the meeting. For example, if you type "sales targets discussed at last month's budget meeting," it will find both the relevant audio segment and the slides that were shared at that time.
Gemma 4 is Google's lightweight AI model announced in April 2026. It has a context window of up to 256,000 tokens (approximately 200,000 characters), allowing it to understand the overall flow even in lengthy meetings.
By combining these two models, Google AI Edge Foresight achieves sophisticated meeting transcription entirely within the Mac.
Google AI Edge Foresight comes with two convenient features.
The first is "real-time note enhancement." If you jot down just "budget increase" during a meeting, the AI analyzes the meeting audio and automatically rewrites it in detail — for example: "Proposal to increase Q2 advertising budget by 15%, from ¥3,000,000 to ¥3,450,000. Approved by Department Head Tanaka."
The second is a "question detection and answering" feature. If someone asks during a meeting, "Where are the minutes from last time?" the AI detects the question, searches past meeting records for relevant information, and displays it on screen.
These features reduce the burden during meetings. You no longer need to focus on taking detailed notes, allowing you to devote your full attention to the discussion itself.
The meeting transcription tool market is already highly competitive, with services such as Otter.ai, Fireflies.ai, and Microsoft Teams' transcription functionality leading the pack.
These services typically cost around ¥1,000 to ¥3,000 per month. Additionally, because they send audio to the cloud, some companies have prohibited their use.
The arrival of Google AI Edge Foresight could change the landscape. If a tool is free and also protects privacy, many business professionals are sure to consider switching.
At present, it is limited to Mac users. No plans for a Windows version have been announced. Also, since it is experimental software, features may change or it could become a paid product depending on future development.
In Japan, meeting culture remains deeply ingrained. Many companies hold meetings multiple times a week and prepare minutes each time.
AI meeting transcription tools have the potential to ease this burden. Many Japanese companies have been cautious about adopting cloud-based tools due to concerns over information leaks.
Fully on-device tools like Google AI Edge Foresight address these concerns. Since data never leaves the company, it should also be easier to obtain approval from information management departments.
Japanese language support accuracy will be a key challenge. At present, usage appears to be centered primarily on English, and no official announcement has been made regarding Japanese transcription accuracy. If Japanese language support is strengthened in future updates, the app could gain widespread adoption in the Japanese market as well.
The emergence of Google AI Edge Foresight demonstrates the potential of on-device AI — AI that runs within the device itself.
In the past, sophisticated AI processing required large-scale servers. With the arrival of lightweight models like EmbeddingGemma 2, it is now possible to achieve sufficient performance on personal computers and smartphones.
This trend will extend beyond meeting transcription to every field imaginable — translation, image editing, coding assistance, and more. An era in which advanced AI capabilities can be used while preserving privacy has arrived.
This article is a cross-post from AI Friends.