Vidita Chandra | 21 May 2025 02:30 AM
At Google I/O 2025, the company unveiled its latest generative AI tools aimed at creators across visuals, video, and music. Leading the charge are Veo 3 for video, Imagen 4 for image generation, and Lyria 2 for music composition—all designed to bring AI deeper into creative workflows. Tying these models together is Flow, a dedicated platform that gives filmmakers a more organised space to craft and refine generative content from idea to final scene.
ALSO SEE: Google I/O 2025: Google’s new tool Flow is an AI filmmaking tool
The most significant leap came with Veo 3, Google’s upgraded text-to-video model. Unlike its predecessor, Veo 3 can now generate native audio—think ambient sounds like traffic or birdsong, and even synthetic dialogue synced to lip movements. It takes both text and image prompts, delivering clips with more realistic physics, sharper detail, and improved temporal consistency.
Meanwhile, Veo 2 hasn’t been left behind—it’s picking up new tricks too. These include reference-based video generation for visual consistency, refined camera controls (like pan and zoom), outpainting to adjust aspect ratios, and tools to add or remove objects with accurate lighting and scale. Several of these updates are now available inside Flow, Google’s new AI-powered filmmaking suite.
For now, Veo 3 is exclusive to Google AI Ultra subscribers in the U.S., accessible via the Gemini app and Vertex AI for enterprise users.
Imagen 4 is Google’s newest image-generation model designed to deliver more refined and visually compelling outputs. The model now supports up to 2K resolution, offering sharper texture details, stronger composition, and notably improved typography—a common shortfall in earlier iterations.
Imagen 4 is already accessible via the Gemini app, Google Workspace tools like Docs, Slides, and Vids, as well as through Whisk and Vertex AI. A faster variant—promising up to 10x the speed of Imagen 3—is on the horizon, aimed at accelerating creative workflows even further.
Google’s latest music-generation model, Lyria 2, is now live in YouTube Shorts and available to enterprise users through Vertex AI. It anchors the company’s Music AI Sandbox, a creative suite that lets musicians compose, tweak, and experiment with entirely new sonic ideas.
For interactive creators, Lyria RealTime brings live performance capabilities—enabling real-time music generation and editing—now accessible via AI Studio. Google has also debuted SynthID Detector, a public-facing tool that allows users to upload media and check whether it was created or modified by AI models using SynthID watermarking.