How Veo is helping the Fukuda Art Museum create “Moving Paintings”

Using technology to explore art in new ways and broaden its accessibility comes with a technical and curatorial challenge: how to move beyond the single, fixed frame. Every static masterpiece, historical photograph, or archival artifact holds not just a captured moment, but the potential for a story that continues beyond the image's edge. The recent initiatives by Google Arts & Culture focused on solving this through Veo — Google's advanced video generation model — creating a framework for animating static visual assets from Moving Archives with Harley Davidson Museum and Moving Paintings with Fukuda Art Museum in Japan, is unlocking new storytelling possibilities for curators.

The technical breakthrough is Veo's ability to extrapolate plausible movement from a fixed composition. The model bridges the gap between a static input and hundreds of video frames, generating temporal coherence that feels intentional.

Google Arts & Culture has developed two distinct operational modes for this process, each designed to answer a different kind of visual question:

1. Animation mode: revealing the narrative

This mode is driven by expert-defined inputs. Curators, in partnership with Google teams, are identifying the implied energy within the scene — the falling rain, the passing traveler, the fluttering banner — to translate these into specific movement vectors. Veo then synthesizes this input to render a continuous, high-definition sequence. The result is a controlled narrative unfolding, transforming the composition's implied story into an explicit visual event, inviting viewers to analyze the moment after the artist's brush left the canvas.

2. Photorealistic mode: imagining the source

This mode addresses the question of what could actually have been here. It focuses purely on contextual and environmental plausibility. Veo uses the static image as a visual seed to generate a high-fidelity video that simulates the photorealistic world that might have inspired the original view. This computer vision process predicts a stable, temporally coherent environment from a single-frame cue, essentially offering a digital window into the reality that preceded the artistic interpretation.

With the help of Veo, we have prototyped a pathway to augment digital archives into dynamic, analysis-ready assets, forging a new pathway for both preservation and visual storytelling.

Explore Moving Paintings at goo.gle/moving-paintings on Google Arts & Culture.