Google AI Studio can start an XR app from a prompt
AI Studio added XR Blocks so a prompt can start an interactive 3D app with hand tracking, tabletop physics, and holograms from text or photos.
Every Google AI Studio update in one feed — release notes, features and changes, straight from the source.
16 updates
AI Studio added XR Blocks so a prompt can start an interactive 3D app with hand tracking, tabletop physics, and holograms from text or photos.
Google AI Studio adds an in-app preview of the Gemini API docs, so developers can reference them without switching tabs.
Google adds token budget caps, resumable interactions, and scheduled recurring runs to the managed agents API.
Replaces the machine-generated deploy link with a permanent, memorable address for apps built with Gemini in AI Studio.
The new model generates 10-second 720p clips and edits them via prompts, priced at $0.10 per second of output.
The new button generates and applies alternate layouts when a text prompt can't capture the look you want.
Google says over 99% came from people who had never built an Android app before, no coding required.
A phone app for building with Gemini is coming soon to app stores, though Google hasn't set a release date.
Apps built in Google AI Studio can now pull in Docs, Sheets and Gmail, be tested on Android, and publish two apps to Cloud Run without a credit card.
Developers can now send an AI Studio project straight into Antigravity, Google's local dev platform, for continued iteration.
Google AI Studio now provisions the sandbox per API call, with agent instructions, skills and tools defined in Markdown.
Google AI Studio will add mid-flight steering and async tool calls on 8 June 2026, alongside breaking changes for existing integrations.
Logan Kilpatrick says the update speeds up building apps from prompts and cleans up generation in Google's free IDE.
The visual edit tool also gets a redesign for apps built in Google AI Studio.
Vibe Coding now shows several Gemini-generated theme options up front, instead of a blank screen during generation.
The text-to-speech model adds inline audio tags, multi-speaker dialogue and 70+ languages, with SynthID watermarking on all output.