Google I/O 2025 showcased major AI advancements, including updates to Gemini 2.5 models (Pro and Flash), Imagen 4, and Veo 3. Gemini 2.5 offers extended language support, improved reasoning, and a new Deep Think mode for complex tasks. Imagen 4 enhances image generation with 2K resolution and better text accuracy, while Veo 3 introduces advanced text-to-video capabilities. Google also launched Flow, an AI filmmaking tool that integrates Imagen, Veo, and Gemini for creating detailed cinematic scenes from text prompts. Flow is now available to Google AI Pro and Ultra subscribers in the US, enhancing creative video production.
Google I/O 2025 was a significant event, showcasing numerous announcements primarily centered around AI. Updates were introduced for the Gemini 2.5 models (Pro and Flash), alongside the enhanced Imagen 4 image generator and the Veo 3 AI video generator. Google also presented Flow—a dedicated AI filmmaking tool that merges Imagen, Veo, and Gemini models to craft cinematic scenes from simple text inputs.
Gemini 2.5 introduces enhanced language support for over 24 languages with advanced text-to-speech and expressive voice options. Google asserts improvements in reasoning, multimodal capabilities, coding, and long context processing for both Flash and Pro models.
A notable addition to the Pro model is the Deep Think reasoning mode, designed for tackling complex math and coding challenges. This feature remains in the “experimental” stage and is set to be shared with trusted testers soon. Google highlights its ability to assess multiple hypotheses prior to responding.
Gemini 2.5 Pro has emerged as the leading AI model on WebDev Arena and LMArena benchmarks, offering superior coding and web app development tools along with a 1 million token window for enhanced long context understanding. Furthermore, Gemini 2.5 enhances protection against indirect prompt injections.
Gemini 2.5 Flash is now available in preview for all users through the Gemini app, with a general release planned for later in June. The commercial launch of Gemini 2.5 Pro will follow shortly thereafter.
Imagen 4 is now capable of generating images at resolutions up to 2K. Google claims enhancements in text accuracy for created cards, posters, and comics, with Imagen 4 available starting today across the Gemini, Google Workspace, Whisk, and Vertex AI applications.
Imagen 4 samples (shared by Google)
Veo 3 is Google’s most recent AI video model, featuring enhanced recognition for text-to-video prompts. It can generate videos with sound, character dialogue, and ambient noises. Veo 3 is now available for Google AI Ultra subscribers in the US as well as Vertex AI enterprise users.
Veo 2 is receiving updates that include camera movements along with capabilities for adding and removing objects. Users can also upload images for style adjustments and outpainting to extend frames beyond their original limits.
Flow is Google’s innovative AI filmmaking tool, which integrates the functionalities of Veo, Imagen, and Lyria models to create cinematic scenes with improved detail. Google asserts that Flow will assist storytellers in crafting impressive cinematic clips that excel in physics and realism. Users can manipulate camera motion, angles, and perspectives, and also edit and extend previously generated videos. Google Flow is currently available to Google AI Pro and Ultra subscribers in the US.