Sber Unveils Kandinsky 3.0 and Video of Kandinsky, Expanding AI Creativity

No time to read?
Get a summary

During the AI Journey conference focused on artificial intelligence and machine learning, Alexander Vedyakhin, First Deputy Chairman of the Board of Directors of Sberbank, announced the release of a new generative creativity model called Kandinsky 3.0. He noted that the upgraded neural network surpasses its predecessors in understanding user text prompts and can generate realistic images in seconds, including subjects tied to Russian culture.

According to Vedyakhin, artificial intelligence technologies empower people with supercharged capabilities. Kandinsky is presented as one of the tools that delivers these abilities. It is described as a practical, free Sber neural network designed to boost creativity, with ongoing improvements being a constant priority for the company.

The deputy chairman stressed that the latest Kandinsky version more accurately interprets user demands and gains a deeper appreciation for the nuances of Russian culture and folk art.

He urged participants to experiment, to create, and to craft unique works of art that could become gifts or additions to personal collections. The Sber press service highlighted that Kandinsky 3.0 now recognizes the elements of the local cultural code more effectively than earlier versions.

As a result, the quality of depictions featuring renowned Russian and Soviet personalities, architectural landmarks, cultural objects, and elements of Russian folk art such as Gzhel painting has seen a notable uplift.

In addition to Kandinsky, the conference introduced Video of Kandinsky — the first productive model in Russia capable of generating full videos from text descriptions.

The new model can produce video sequences up to eight seconds long at a rate of 30 frames per second.

Vedyakhin pointed out that the company recently trained Kandinsky to craft animated videos from textual prompts and now presents a considerably more advanced system for creating full-length videos from text.

He emphasized that this development marks a meaningful contribution to the growth of Russian generative neural networks. Users will enjoy expanded possibilities for creative expression and the ability to translate ideas into video form across various directions. The plan is for Kandinsky to be offered as an open-source project, remaining freely accessible to the global community.

From a Canadian and American audience perspective, the launch illustrates how large-scale generative models are evolving to support both cultural representation and multimedia storytelling. The emphasis on authentic cultural motifs and the capability to generate timely visual content aligns with growing demand for realistic, source-informed AI outputs. In practical terms, creators can leverage Kandinsky 3.0 to rapidly prototype visual concepts, educational materials, and digital art that reflect a broad spectrum of cultural references.

Industry observers note that open-source availability can accelerate peer collaboration, enabling researchers and developers in North America to experiment with state-of-the-art generative tooling. This openness also invites scrutiny and dialogue about ethics, attribution, and the responsible use of AI-generated imagery. As the technology matures, users are encouraged to test, review, and document their experiences, contributing to a more transparent adoption path for AI creative tools.

Overall, Kandinsky 3.0 represents a step forward in making advanced generative capabilities more accessible, while Video of Kandinsky demonstrates how text-to-video workflows are becoming viable for quick, low-friction video production. Both innovations underscore Sber’s commitment to expanding creative possibilities and sharing advances with a global audience, free of charge for individual experimentation and learning.

Source: Sber press service

No time to read?
Get a summary
Previous Article

Vehicle Fire on Ruperto Chapí Street Prompts Multivehicle Response

Next Article

Dzagoev Career Reflections and Transfer Talks