Google’s Omni Flash turns videos into AI-generated scenes

A plush deer called Buddy is back on screen, this time in tests of Google’s Omni Flash video model. The tool can take an uploaded video and a text prompt to create a new video, and Google says it is better at using real-world knowledge and keeping characters consistent than its predecessor, Veo.

Corp and Tech·November 5, 2026·reads: 1,225

Google’s Omni Flash turns videos into AI-generated scenes

Omni is a new family of generative models that Google says is intended to transform different kinds of input—such as photos, video and text—into other formats. Its first release, Omni Flash, generates video and is available through Flow, Google’s AI video generation and editing platform. Veo remains available as an alternative.

To test the claims, the reviewer returned to Buddy, a child’s stuffed deer previously deepfaked into vacation scenes while the reviewer tried to recreate an advert by Google. The videos were never shown to the reviewer’s four-year-old. The earlier experiment had highlighted how little effort and technical knowledge realistic AI video can require, as well as the uncertain boundary between harmless fun and “slop.”

The reviewer used Buddy again to see how Omni Flash handled a new adventure. The results, according to the source, were a mixed bag—baffling rather than a clear demonstration that the model’s claimed improvements work consistently.