Google is testing Gemini Live on desktop and web platforms, potentially bringing the real-time voice mode beyond its current mobile-only availability.

The voice feature remains absent from Google's desktop app despite months of development signals. The company promised new voice capabilities for the macOS app at I/O in May, with earlier builds showing voice selection options and screen-sharing overlays.

Testing is reportedly underway through Google's Trusted Tester program, with a simultaneous launch across desktop and web appearing likely.

Skills expansion beyond Spark

Google is also preparing to expand Skills functionality into regular Gemini chats. Skills currently exist only within Gemini Spark, the autonomous agent available to AI Ultra subscribers, where they function as reusable instruction packages for recurring tasks.

New builds suggest ordinary chat users could soon upload their own skills, select from predefined options, build new ones from scratch, or have Gemini generate skills automatically.

This would close a notable feature gap with Claude and ChatGPT, both of which already offer similar capabilities. Chrome received a lighter prompt-shortcut version of skills in April, while Gemini Enterprise already allows workers to invoke skills mid-chat.

Timing remains uncertain

The rollout timeline remains unclear. Gemini 3.6 Flash appears to be in preparation while Gemini 3.5 Pro has missed multiple targets, most recently a mid-July deadline.

A stopgap model could arrive first, with potential announcements by the end of July. However, Google has not confirmed these features, and development plans can shift before launch.

Together, Live on web and chat-level Skills would significantly enhance the everyday Gemini experience, bringing it closer to parity with competing AI assistants.