NEvo: Neural-Guided Evolutionary Video Synthesis
What kind of video would make a given part of your visual brain light up the most? NEvo answers this in silico: it evolves AI-generated videos to maximally drive a target brain region, then uses them to map how visual selectivity becomes increasingly dynamic and social along the lateral stream. EXPLORE THE ATLAS Hover a region on the cortical surface to reveal the stimuli synthesized to drive it. ACROSS THE SURFACE A dense searchlight across the surface, each covered sphere has its own synthesized stimuli. THE METHOD We train an encoding model, a “digital twin”, that predicts how each visual region responds to any video. Now we can ask: which video would make a chosen region light up the most? NEvo searches for that video automatically, using the twin’s prediction as its reward. Every video is described by a handful of "genes", subject, lighting, motion, mood. NEvo treats them like a population: generate a batch, score each with the digital twin, then keep, mix (crossover) and tweak (m