- Core update bright spots
- Gemini 2.5 Flash supported Native Voice Dialogue (Live API):
- # Text-to-Speech (TTS) upgrade:
- 1. New “Build” tab: extremely simple generation from prompt to App
- 2. ** Code helper functionality online: support version comparison and withdrawal**
- 3. ** Key deployed to Cloud Run: zero-wire AI App Production**
- 4. New “Generate Media” page: integration of all multi-model model capabilities
- 5. ** Support for MCP standards: Enhanced Ecocompatibility**
- 6. ** Experimental function: URL Context, let model read web page**
Google officially launched a new upgrade on I/O 2025** for the developers to provide a stronger AI primary development platform. This update significantly enhanced the usability, integration and ease of deployment of the Gemini model by creating the “** one-stop AI App Build Platform** “ .
Core update bright spots
1. Native Code Generation
-
Introduction of a code editor Gemini 2.5 Pro to Studio to support the direct generation of text/image/video tips for Web applications;
-
Add Build Tab: Accelerant-based web application (support to deployment of one key to Claude Run) can be built and deployed quickly;
-
Support for iterative development: you can change the code, see the diff, return the historical version in the chat dialogue.
##2. The Media Generation Centre (Generate Media)
-
Integration Imogen (image),Veo (video),** Lyria (music)** and Gemini (text) multi-model capacity;
-
New interactive music generation App: PromptDJ, based on Lyria RealTime.
##3.** Audio capability upgrade: more natural, more responsive**
Gemini 2.5 Flash supported Native Voice Dialogue (Live API):
-
Support for 30 multiple natural persons (male and female, accent, emotional)
-
Introduction of ** active audio recognition**: a model that distinguishes between a user ‘ s speech and background noise and responds only at the appropriate time;
-
Fits more closely to the rhythm of human natural dialogue and applies to such scenes as customer service, virtual assistants, interactive dramas, etc.
# Text-to-Speech (TTS) upgrade:
-
Single-person or multi-person dialogue generation;
-
Support for multidimensional control of speech speed, tone and emotions;
#Agentic and Tool Ecology
-
** Add a new `Build’ tab** to create a Gemini application prototype from text, image or video prompt to integrate the Gemini 2.5 Pro model.
-
** Smart Code Assistant** supports the editing of existing application codes, viewing differences (diff), rollback history versions.
-
Step one key to Cloud Run to automatically host Gemini API Key to the production environment without the need to configure the server.
-
The new “Generate Media” page focuses on multi-model models such as Imogen (image), Veo (video), voice generation, etc.
-
** Supports the MCP (Model Context Protocol)** primary integrated open source standard, which facilitates the construction of complex AI application and docking third-party tools.
-
URL Context experimental function Models can read the content of web links for summary, match, research and verification.
1. New “Build” tab: extremely simple generation from prompt to App
-
Closely co-ordinated with Google’s Genai SDK
-
Supports the direct generation of application prototypes from text, images or videos
-
Auto-generated UI+ functional code (fit for front end or full-store prototype)
2. ** Code helper functionality online: support version comparison and withdrawal**
-
Provide “see change” capability
-
Support back to the historical version (checkpoint mechanism)
-
Significant improvements in multi-person collaboration and version control efficiency
3. ** Key deployed to Cloud Run: zero-wire AI App Production**
-
Upon completion of the application build, direct deployment to Google Cloud Run
-
Gemini API Key will automatically save it on the service end and improve security.
-
Enable developers to move prototypes to the online environment for use
4. New “Generate Media” page: integration of all multi-model model capabilities
-
Centralized access and use: Imogen (image generation)
-Native speech model (voice generation)
One-stop call multi-module modeling, adapted to applications of creativity, education, content, etc.
5. ** Support for MCP standards: Enhanced Ecocompatibility**
-
Provision of standard interfaces for the construction of complex dialogue systems or multi-model interactive applications
Example screenshot display: Colosseum query side by side with code view
6. ** Experimental function: URL Context, let model read web page**
-
Executive summary
-
Information comparison.
Details: https://develators.googleblog.com/en/google-ai-studio-native-code-generation-agents-upgrade/