Omnimattezero tool eliminates need for costly AI training, making video layer editing/ background separation faster, cheaper, more accessible

https://www.eurekalert.org/news-releases/1100089

"extracts objects/ figures from backgrounds, preserving complex elements (fur, hair, foliage, shadows, reflections, smoke, rippling water)... enables content recycling, leveraging image-completion enhanced with modules tracking changes across time/ space maintaining reconstructed background consistency through built-in self-attention mechanism linking regions across/ within video frames... existing video-generation models detect object-created effects, remove/ extract/ reinsert them into other videos in real time"