.jpg "UK data regulator to move to Manchester"){kind=logo}
Google announced the rollout of the Veo 2 artificial intelligence (AI) model to Gemini chatbot on Tuesday. Currently, the AI model will only be accessible to the paid subscribers of Gemini. With this AI model, eligible users can generate eight-second-long videos with text prompts in natural language. The Mountain View-based tech giant first introduced Veo 2 in December 2024 as the successor to the Veo AI model. It is also available via the company’s Vertex AI platform, and it powers YouTube’s Dream Screen feature.
Veo 2 in Gemini Can Generate 720p Videos
In a blog post, the tech giant announced the rollout of the video generation AI model in Gemini. Google is rolling it out globally in all the languages Gemini supports to the Gemini Advanced subscribers. Those on the free tier will not have access to the model.
Veo 2 can be selected from the model picker menu in both Gemini’s web client and mobile apps. The model can generate an eight-second video clip based on text prompts. The output video is generated in 720p resolution with a 16:9 aspect ratio and can be downloaded in MP4 format. Google says there is a monthly limit on how many videos a user can generate, and it will notify the user as they approach the limit.
Google is also letting users share the generated videos with others. On the Gemini app, users can tap the share button to quickly upload videos to social media platforms such as TikTok and YouTube. Notably, while the feature has begun rolling out, it may take a few weeks before eligible users across the world have access to it.
At the time of launch, the tech giant said that Veo 2 comes with improvements in realism and details of the various elements. It can also understand the technical terms of cinematography, which means users can specify a camera lens, camera movements, cinematic effects, and more in text prompts, and the AI model can adhere to them.
Google also claimed that the video generation model has been trained to hallucinate less. As a result, users should see unwanted details such as extra fingers, morphed limbs, or unexpected objects less frequently.
