Google I/O 2025 Keynote highlights

The Google I/O 2025 Keynote was held today as the event saw a host of announcements focused on AI. For those that missed it, Google shared updates and new features for the Gemini 2.5 models (Pro and Flash).

Other notable highlights included the Imagen 4 image generator, Veo 3 AI video generator, and Flow—a dedicated AI-powered filmmaking tool.

Gemini 2.5

The Gemini update adds improvements to text-to-speech and voices with extended language support to over 24 languages. The company says among its updates are improved reasoning, multimodality, coding, and long context capabilities for Flash and Pro models.

The Pro model is a recipient of a huge update, adding Deep Think reasoning mode for complex math and coding tasks. However, this feature is still in its experimental phase and is to be released to testers soon.

As per Google, it’s capable of considering multiple hypotheses before making a response. It also improves security against indirect prompt injections.

Meanwhile, Gemini 2.5 Flash is available for preview to all users with the Gemini app. Its general release isn’t due until later in June, with Gemini 2.5 Pro’s commercial release following shortly after.

Imagen 4

Imagen 4 can now produce images with up to 2K resolution, offering improvements in text accuracy for generated cards, posters, and comics.

For those curious, it’s available today across the Gemini, Google Workspace, Whisk, and Vertex AI apps. Here are some samples below.

Veo 3

Veo 3 is Google’s latest AI video model that features some appreciated changes to text-to-video prompt recognition. The model can output video with sound, character dialogue, and background noises.

It’s available today for Google AI Ultra subscribers in the US and Vertex AI enterprise users. In addition, Veo 2 has new features like camera movement, object addition,and removal functionality.

Users may also add images for more creative flexibility and outpainting to extend frames beyond their original borders.

Flow

Lastly, we have Flow, which combines the Imagen and Lyria models to create cinematic scenes from simple text prompts. The company claims that this program can help storytellers produce clips that excel at physics and realism.

Users have the ability to control camera motion, angles, and perspectives. You may also extend and edit previously generated videos. For now though, it’s only available to Google AI Pro and Ultra subscribers in the US.