At its annual developer conference, Google I/O 2025, the company introduced generative AI tools focused on image, video, and music creation. The new models, Veo 3 for video, Imagen 4 for images, and Lyria 2 for music, integrate AI into creative workflows, while the new tool Flow offers a more structured environment for filmmakers working with generative content.
Also Read: Gemini Live is rolling out on Android and iOS
Veo 3 Supports Native Audio Generation
The biggest upgrade came in the form of Veo 3, Google’s latest text-to-video model. Unlike its predecessor, Veo 3 supports native audio generation, meaning it can generate environmental sounds, such as traffic or birds, and even synthetic dialogue that matches lip movement. The model can process both text and image prompts and produces videos with more accurate physics, better realism, and improved temporal consistency.
For now, Veo 3 is limited to Google AI Ultra subscribers in the U.S., available through the Gemini app and via Vertex AI for enterprise users. Google says the model is already being used in professional workflows, with filmmakers helping shape its development.
Alongside Veo 3, Google has also added new features to Veo 2, including:
- Reference-powered video generation (to maintain visual consistency),
- Camera controls (for panning, zooming, and other movements)
- Outpainting (to change aspect ratios), and
- Object addition/removal from scenes with accurate scaling and lighting.
- Some of these tools are now integrated into Flow, Google’s new AI filmmaking platform.
Flow: A Filmmaking Tool Built Around AI Models
Flow is Google’s new environment for creating AI-generated videos using Veo, Imagen, and Gemini. It allows users to generate and manage scenes, characters, and objects using text prompts or custom image inputs. The tool also includes asset management features, camera movement controls, and a scene builder to refine transitions and maintain continuity across shots.
Flow is currently available to Google AI Pro and Ultra subscribers in the U.S., with plans to expand to other countries.
Flow is also being tested with a small group of professional filmmakers. One of the key collaborations is with director Darren Aronofsky, whose studio Primordial Soup is partnering with Google DeepMind to explore the role of generative models in short-form storytelling. The first film from this effort, Ancestra, will debut at the Tribeca Film Festival in June.
Imagen 4 Brings Improved Detail and Typography
Google also introduced Imagen 4, its latest image-generation model. It produces higher-resolution images (up to 2K), with improvements in texture detail, composition, and most notably, typography. According to Google, Imagen 4 is better at handling text inside images, a known weakness in previous models.
Imagen 4 is available through the Gemini app, Workspace tools (Docs, Slides, Vids), Whisk, and Vertex AI. A faster version of Imagen 4, claimed to be up to 10 times quicker than Imagen 3, is expected soon.
Lyria 2 and RealTime: Music Tools for Composers
Lyria 2, Google’s music-generation model, is now available in YouTube Shorts and to enterprise users via Vertex AI. It’s part of Google’s Music AI Sandbox, which provides musicians with tools for composing, editing, and experimenting with new sounds.
In addition, Lyria RealTime allows users to create and modify music interactively, including live performances, and is now accessible through AI Studio.
Google also launched SynthID Detector, a public tool that allows users to upload content and check whether it was created or altered by AI models using SynthID.
Read More: Google takes on Meta with stylish, AI-powered Android XR glasses
Follow Us
Siddharth Chauhan
Siddharth reports on gadgets, technology and you will occasionally find him testing the latest smartphones at Digit. However, his love affair with tech and futurism extends way beyond, at the intersection of technology and culture. View Full Profile
