Google Gemini Desktop App Aims to Catch Up with Web Version
According to ComputerBase, citing data from Testing Catalog, Google is making efforts to align the functionality of the Google Gemini desktop app with its web version. Recent test builds show several key features that were previously available only through the browser.
Image and Video Generation, Plus Direct Camera Access
The new navigation includes dedicated tabs for image and video generation. This suggests Google plans to bring functions, previously accessible mainly via the web interface, directly into the desktop app.
A separate block for camera work has also appeared. The menu now includes an option to launch recording mode. When selected, it opens an interface for creating individual photos that can then be immediately inserted into a prompt (request) for the model. It is important to note: this refers to static images, not a continuous video stream. This distinguishes the feature from Gemini Live, which is designed for processing real-time video streams.
Enhanced Spark and MCP Server Support
The third innovation is support for more flexible configuration of the Spark agent. The connectors menu now allows adding custom MCP (Model Context Protocol) servers. Previously, such configuration was possible only through hidden internal mechanisms, not via the standard user interface. This is particularly relevant for Mac users, as it allows connecting external tools and data sources more conveniently directly within the desktop app environment.
Development Status
These updates follow a series of expansions released in July. The Spark agent is already available on macOS (with integration with Google Tasks and Keep), as well as via Chrome or Google Workspace on other platforms. Since July 29, intelligent dictation and screen context analysis were added to the functionality.
With the addition of new blocks for multimedia and camera work, Google continues to bridge the gap between the desktop and web versions, so users rarely need to switch to the browser. However, exact release dates for these features for the general audience are still unknown. Access to the corresponding builds is currently limited to selected testers.