At the 2025 Google I/O Developer Conference, Google showcased its latest achievements in AI and XR technologies, covering three core areas: creation tools, developer platforms, and user experience. The next-generation video generator Veo3, image generator Imagen4, brand new XR platform Android XR, and top-tier AI subscription service "AI Ultra" became the highlights of the conference.
As a major upgrade to AI creation tools, Veo3 can generate native audio-video content, distinguishing itself from current products like OpenAI Sora that do not support audio. It is now available to U.S. Ultra subscribers and Vertex AI enterprise customers. Imagen4 offers improved layout and image details while supporting multi-format outputs. Both have been integrated into the Gemini app, Workspace, Whisk, and the new AI film-making platform Flow.
Flow combines scene generation, lens control, and material management functions, allowing users to blend AI-generated content with original works to create full-scale cinematic workflows. Meanwhile, Google has also released the music generation model Lyria, providing AI creation capabilities to music creators through YouTube Shorts and Vertex AI.
To meet the needs of high-end creators and enterprises, Google launched the AI Ultra subscription at $249.99 per month, offering top-tier services including Veo3, Imagen4, Project Mariner proxy tools, YouTube Premium, and 30TB cloud storage. There's also a Pro version as a secondary option.
Gemini Live will soon be available on the Gemini app, supporting real-time collaboration via camera and screen sharing, suitable for troubleshooting and event planning. Gemini functionalities are also embedded in Chrome for web content analysis and summarization.
In terms of communication, Google Meet will introduce an AI translation feature that mimics tone and expression, initially supporting English and Spanish, available only to Pro and Ultra users. Gmail will also introduce personalized smart replies based on users' historical email data, requiring user authorization to enable.
On the developer side, the Gemini API adds advanced features such as asynchronous functions, native audio, and tool context protocols (MCP). New models like Gemma3n (lightweight mobile model), **MedGemma (medical applications), and SignGemma (sign language recognition)** further expand Google's AI model landscape.
To enhance the credibility of AI-generated content, Google also launched the SynthID Detector portal, which identifies AI-generated sources of images, text, audio, and video through invisible watermarks. Over a billion projects have used this marking system.
Finally, Google announced the launch of a new AR/MR platform, Android XR, to provide operating system support for future XR devices, including AR glasses and headsets. Combined with Gemini intelligence, the device will support real-time translation, environmental analysis, message handling, and more. Google has partnered with brands like Gentle Monster and Warby Parker and plans to release the SDK this year.
This conference indicates that Google is deeply integrating AI capabilities into its products, content creation, and end-user platforms, fully entering the construction of an AI-native ecosystem.