Earlier in the I/O 2025 event,Google They also released the next-generation Imagen 4 image generation model and Veo 3 video generation model. Not only has the visual quality greatly improved, but generation speed has also become faster, especially for images, reportedly up to 10 times faster.
Google launches the next-generation Imagen 4 image generation model.
Imagen 4 is currently Google’s newest and most powerful image generation model, bringing significant improvements, such as richer image textures, more refined color gradations, and greatly enhanced detail rendering, including changes in light and shadow, the texture of water droplets, and more.
Especially for text, Google emphasized at the launch event that this model has “better support for text and layout,” and it’s not just about generating words correctly but also enabling creative design—for example, using dinosaur bone elements as a typeface in a music festival poster to make the poster more dynamic. However, how well it supports Chinese characters remains to be tested:
Compared to the previous generation, Imagen 4 generates 10 times faster, which is quite nice. When you’re not satisfied with the generated image, you can quickly make changes without worrying about waiting too long. Combined with Gemini’s editing capabilities, you can even modify small parts:
Google’s new-generation Veo 3 video generation model debuts
Veo 3 is Google’s latest next-generation video generation model, featuring significant upgrades compared to its predecessor Veo 2.
Now Veo3 can not only produce high-quality videos, but also simultaneously generate sound effects, ambient sounds, and dialogue, meaning that once the video is generated, the characters in it are also fully voiced, and there are even background sounds:
The image details, lighting and shadow, and physical behaviors (such as water ripples, wind movement, etc.) are also naturally more lifelike, delivering an immersive visual experience. Google also emphasized that Veo 3 has stronger physical understanding, allowing it to handle dynamic objects and environmental interactions more naturally. Veo 3 is now available starting today.
Flow creation tool that combines Imagen 4, Veo 3, and Gemini
To make it easier for filmmakers to use Google’s various AI generation tools, this release also introduces a new Flow tool that combines Imagen 4, Veo 3, and Gemini capabilities, enabling users to generate an outstanding piece through natural language:
Google also gives an example of a workflow, such as making a video of a grandfather and a chicken driving a flying car. First, upload pictures of the old man and the car, but since he doesn’t have any chicken material on hand, he uses Imagen to generate it.
Then enter the prompt to let Veo integrate all the materials and generate a short film:
For the video’s continuation, if you don’t know how to narrate it, you can also let the AI judge and generate it on its own:
Of course, the generated videos can be exported and downloaded, then imported into any editing software you commonly use for post-production.
All of the above actions are completed within the Flow platform, with no need to switch back and forth.
For this Flow launch, Google also announced two new paid subscription plans, namely “Google AI Pro,” at $19.99 per month, which grants access to Flow and Veo 2:
Another option, “Google AI Ultra,” is rolling out globally for $249.99 per month, granting access to Flow and Veo 3, along with the Deep Think feature of Gemini App 2.5 Pro. It also includes YouTube Premium, 30TB of storage, Project Mariner, and more:
Source: KOCPC Chinese