Google AI Studio 在 Google I/O 大会上重磅发布一系列更新，旨在改善开发者使用 Gemini 和其他模型进行构建的体验。主要功能包括编辑器原生集成的代码生成，利用 Gemini 2.5 Pro 的编码能力，通过文本或图像提示快速创建 Web 应用程序。新的“构建”选项卡助力应用程序快速开发，并支持一键部署至 Cloud Run。更轻松地访问 Imagen、Veo 及原生语音生成等模型，多模态生成能力显著增强。该平台还为 Live API 引入了新的原生音频功能。Gemini 2.5 Flash 预览版支持自然对话和主动音频，同时更新的文本转语音 (TTS) 功能，Gemini 2.5 Pro 和 Flash 预览版提供多说话人输出和样式控制。此外，Google Gen AI SDK 现在支持模型上下文协议 (MCP)，以便更好地与外部工具集成，并且提供了一个实验性的 URL 上下文工具，支持模型从链接中检索信息。值得注意的是，一些功能（如一键代码生成和 URL 上下文工具）被标记为实验性功能。这些更新旨在使 Google AI Studio 成为开发者探索和构建 Google 最新 AI 模型的重要平台。
Native code generation, agentic tools, and more
Google AI Studio is the fastest place to start building with the Gemini API, with access to our most capable models, including Gemini 2.5 preview models, and generative media models like Imagen, Lyria RealTime, and Veo. At Google I/O, we announced new features to help you build and deploy complete applications, new model capabilities, and new features in the Google Gen AI SDK.
Build apps with Gemini 2.5 Pro code generation
Gemini 2.5 Pro is incredible at coding, so we’re excited to bring it to Google AI Studio’s native code editor. It’s tightly optimized with our Gen AI SDK so it’s easiery to generate apps with a simple text, image, or video prompt. The new Build tab is now your gateway to quickly build and deploy AI-powered web apps. We’ve also launched new showcase examples to experiment with new models and more.
In addition to app generation from a single prompt, you can continue to iterate your web app over chat. This allows you to make changes, view diffs, and even jump back to previous checkpoints to revert edits.
You can also deploy those newly created apps in a single click to Cloud Run.
Google AI Studio apps and generated code leverage a unique placeholder API key, allowing Google AI Studio to proxy all Gemini API calls. Consequently, when you share your app with Google AI Studio, all API usage by its users is attributed to their Google AI Studio free of charge usage, completely bypassing your own API key and quota. You can read more in our FAQ.
This feature is experimental, so you should always check code before sharing your project externally. Our one-shot generation has been primarily optimized to work with Gemini and Imagen models, with support for more models and tool calls coming soon.
Multimodal generation made easy in Google AI Studio
We've been working hard to get Google DeepMind’s advanced multimodal models into developers’ toolboxes, faster. The new Generate Media page centralizes the discovery of Imagen, Veo, Gemini with native image generation, and new native speech generation models. Plus, experience interactive music generation with Lyria RealTime with the PromptDJ apps built in Google AI Studio.
Link to Youtube Video (visible only when JS is disabled)
New native audio for the Live API and text-to-speech (TTS)
With Gemini 2.5 Flash native audio dialog in preview in the Live API, the model now generates even more natural responses with support for over 30 voices. We’ve also added proactive audio so the model can distinguish between the speaker and background conversations, so it knows when to respond. This makes it possible for you to build conversational AI agents and experiences that feel more intuitive and natural.
Link to Youtube Video (visible only when JS is disabled)
In addition to the Live API, we’ve announced Gemini 2.5 Pro and Flash previews for text-to-speech (TTS) that support native audio output. Now you can craft single and multi-speaker output with flexible control over delivery style.
Link to Youtube Video (visible only when JS is disabled)
Try native audio in the Live API from the Stream tab and experience new TTS capabilities via Generate Speech.
Model Context Protocol (MCP) support
Model Context Protocol (MCP) definitions are also now natively supported in the Google Gen AI SDK for easier integration with a growing number of open-source tools. We’ve included a demo app that shows how you can use an MCP server within Google AI Studio that combines Google Maps and the Gemini API
New URL Context tool
URL Context is a new experimental tool that gives the model the ability to retrieve and reference content from links you provide. This is helpful for fact-checking, comparison, summarization, and deeper research.
Start building in Google AI Studio
We’re thrilled to bring all of these updates to Google AI Studio, making it the place for developers to explore and build with the latest models Google has to offer.
Explore this announcement and all Google I/O 2025 updates on io.google starting May 22.
