Google has launched a new feature for its Gemini AI assistant.

This feature allows paid subscribers to convert photos into short video clips.

Initially available to a limited audience earlier this year, it became accessible to users of the Google AI Ultra and Pro plans on July 10, 2025, in select regions.

Subscribers can use the feature on the web version of Gemini, with a rollout to the mobile app expected within the week.

The tool enables users to create eight-second video clips in MP4 format at a resolution of 720p, based on a photo and optional text descriptions.

This new feature is powered by Google’s Veo 3 video generation model, which was introduced at the company’s developer conference in May 2025.

Veo 3 is also available through Flow, a separate paid filmmaking tool.

The addition of this feature positions Google to compete with other companies such as OpenAI, Runway AI, and Chinese firms like Alibaba and Kuaishou Technology, which have also released AI-powered video tools.

🔗 Source: Bloomberg

🧠 Food for thought

1️⃣ AI video generation joins the escalating arms race in generative technology

Google’s photo-to-video feature represents the latest development in the rapidly accelerating field of generative AI, where competition is intensifying globally among tech giants.

The feature stands alongside similar innovations from OpenAI, Runway AI, and Chinese competitors like Alibaba and Kuaishou, indicating a worldwide race to dominate this emerging technology 1.

Google’s approach with its Veo 3 model through Gemini follows a pattern similar to OpenAI’s development path, initially releasing powerful generative capabilities to limited audiences before gradually expanding access to paying customers 2.

This competitive pattern mirrors earlier AI development cycles, where companies have historically prioritized capability advancement and market position before fully resolving ethical concerns and technical limitations 3.

The technology sits at the intersection of various AI disciplines that have evolved since the field’s formal establishment at the 1956 Dartmouth Conference, drawing on advances in computer vision, natural language processing, and neural networks 4.