Google is better at video generation again. Veo 3.1 adheres to instructions more precisely and renders materials better

Five months after the release of the original Veo 3, Google introduced a significantly improved Veo 3.1 model. While the first generation of this AI video creation tool arrived in the Czech Republic last July, the new version brings a number of technical enhancements that push the possibilities of AI-powered video generation to the next level. As the model’s name suggests, it’s not a revolution, but the improvement should be noticeable, at least according to Google.

More Faithful Imagery and More Accurate Interpretation of Prompts

The main advantage of the Veo 3.1 model is its significantly improved ability to follow text instructions. While previous versions sometimes ignored details in the prompt, the new model more accurately interprets user requirements and generates videos that better match the original vision. Google has also improved the realism of textures and materials – generated videos now reportedly look more natural and faithfully capture object surfaces.

A significant new feature is integrated audio support. The model can generate not only visuals but also a corresponding audio track that complements the visual content. This feature is currently experimental, and Google is continuously improving it based on user feedback.

Improved Conversion of Static Images to Video

Veo 3.1 brings improved audiovisual quality when converting images to videos. The feature that allows animating static images now produces smoother and more realistic results. The model better maintains object consistency during movement and can animate complex scenes more naturally.

Google uses the term “state-of-the-art” in connection with the new model, indicating that the model achieves top results in video generation benchmarks. The improvements are based on the original Veo 3 architecture, which engineers expanded with new capabilities while retaining its core strengths.

Availability for Developers and Businesses

The Veo 3.1 model is immediately available in three ways. Regular users can try out the new capabilities directly in the Gemini app. Developers gain access to it via the Gemini API, where they can integrate the model into their own applications. Enterprise customers can then use Vertex AI for deployment in business solutions.

Do you plan to try out the new video generation capabilities with the Veo 3.1 model?

Source: Google