Google AI Studio is the fastest way to build with the Gemini API, giving you access to powerful models like Gemini 2.5 and generative media models such as Imagen, Lyria RealTime, and Veo. At Google I/O, we announced new features to help you build and deploy complete applications, along with updates to the Google Gen AI SDK.
Thank you for reading this post, don't forget to subscribe!
Enhanced Code Generation with Gemini 2.5 Pro
Gemini 2.5 Pro, now available in Google AI Studio’s native code editor, excels at coding. The new Build tab makes it easy to generate and deploy AI-powered web apps from a simple text, image, or video prompt. You can also:
- Iterate apps via chat: Make changes, view differences, and revert to previous versions.
- Deploy instantly: Publish your new app to Cloud Run with a single click.
Apps and code generated within Google AI Studio use a unique placeholder API key. This means any API usage by users of your shared app is attributed to their free Google AI Studio quota, not yours. While this is an experimental feature, it’s a great way to let others try your creations without impacting your own usage limits. Always review the code before sharing it externally.
Multimodal Generation Made Simple
The new Generate Media page centralizes discovery of our latest multimodal models, including:
- Imagen for image generation.
- Veo for video.
- Gemini with native image generation.
- New native speech generation models.
You can also try PromptDJ, an app built in Google AI Studio that lets you experiment with interactive music generation using Lyria RealTime.
New Audio Capabilities
We’ve added new audio features to make conversational AI more natural:
- Live API: Gemini 2.5 Flash now offers native audio with over 30 voices. It uses proactive audio to distinguish between the speaker and background conversations, helping the model know when to respond.
- Text-to-Speech (TTS): Gemini 2.5 Pro and Flash now support native audio output for TTS. You can create single or multi-speaker output with flexible control over the delivery style.
Tooling and Integrations
- Model Context Protocol (MCP): The Google Gen AI SDK now natively supports MCP definitions, making it easier to integrate with open-source tools. We have a demo app that shows how to use an MCP server to combine Google Maps with the Gemini API.
- URL Context tool: This new experimental tool allows the model to retrieve and reference content from links you provide. This is useful for fact-checking, summarizing, and deeper research.
Advanced Tooling and Agentic Capabilities
Google AI Studio is making it easier to build sophisticated, multi-tool AI agents. The platform now offers native support for Model Context Protocol (MCP) definitions within the Google Gen AI SDK. This is a significant step towards creating agents that can seamlessly integrate with a growing number of open-source tools. A new demo application is available to showcase how an MCP server can be used to combine the Google Maps API with the Gemini API, providing a practical example for developers to follow.
URL Context for Enhanced Information Retrieval
A new experimental tool called URL Context allows models to retrieve and reference information directly from provided links. This feature is particularly useful for tasks that require real-time or specific data, such as fact-checking, comparison analysis, summarization, and deeper research. By giving the model the ability to access and understand content from external URLs, it can provide more accurate and context-aware responses.
Live API Improvements for Conversational AI
The Live API has received a major upgrade with the preview of Gemini 2.5 Flash native audio dialog. This allows the model to generate more natural and expressive responses with support for over 30 voices. Additionally, a new proactive audio feature enables the model to distinguish between a speaker and background conversations, helping it determine the appropriate time to respond. These improvements are designed to help developers create more intuitive and lifelike conversational AI agents.
Seamless Deployment to Cloud Run
For developers who are ready to deploy their creations, Google AI Studio now offers one-click deployment to Cloud Run. This makes the transition from development to production incredibly smooth, allowing you to quickly share your AI-powered applications with others. The generated apps use a unique placeholder API key, which means that any API usage by users of your application is attributed to their own Google AI Studio quota, not yours. This provides a worry-free way to share your projects.