Google Now Lets Mac Users Talk To Gemini In Virtually Any App
First previewed at Google I/O, the official release arrives in app version 1.88 and alters how Mac users will interact with the assistant. Instead of toggling back and forth between separate windows, copying text, or keeping a dedicated browser tab open, long-pressing the Fn key summons a sleek floating pill, housing the recognizable Gemini audio waveform right at the bottom of your screen.
Because the shortcut works system-wide, the floating bar overlay
registers voice input over third-party creative suites, word processors,
and file managers alike, letting users voice trigger AI events without losing their current cursor focus or workflow placement.
This update focuses on two primary capabilities: advanced speech-to-text transcription and opt-in desktop screen reasoning. The dictation feature functions similarly to Google's Gboard Rambler engine, inserting polished text directly wherever your text cursor happens to be resting. Rather than transcribing every verbal misstep verbatim, Gemini automatically strips out filler sounds like "umm" and "ah" while accounting for mid-sentence corrections on the fly. If you rephrase a sentence halfway through speaking, the system catches the adjustment and outputs clean, ready-to-use copy across any native macOS application.
The opt-in context mode gives Gemini the ability to read and process visual information actively displayed on your monitor. Once manually enabled within the app, users can highlight local desktop files, documents, or images and issue verbal requests. For instance, highlighting a folder of medical records lets you instruct Gemini to parse the documents and draft a concise summary directly into an email body.
This desktop awareness extends seamlessly into live text editing and creative asset management as well. You can highlight messy, unformatted meeting notes anywhere on your screen and verbally request an executive summary complete with a bulleted overview at the top. On the visual side, Gemini can generate new images or iterate on existing desktop graphics through conversational prompts, such as referencing a light-mode vector illustration on your screen and asking the assistant to produce a dark-mode variation.
The rollout is currently expanding globally to English-language setups (which means you may not see if right away even on v1.88), with additional languages planned for upcoming releases.
