Simplify How-it-works, drop gallery caption column, halve gallery videos

README.md

@@ -28,69 +28,24 @@ It orchestrates three frozen models (weights are **not** bundled — they are pu
 
 Each clip is the **top result of a NEvo search targeting one visual region** — the model discovers, from scratch, stimuli that drive that region's known selectivity.
 
-| Region | Synthesized stimulus | What the search converged on |
-|--------|:--------------------:|------------------------------|
-| **FFA** · faces | ![FFA](assets/gallery/FFA.gif) | *fashion editorial, multiple faces, crowd of people* |
-| **PPA** · places | ![PPA](assets/gallery/PPA.gif) | *unreal-engine render, subway station, looking-up view* |
-| **MT** · motion | ![MT](assets/gallery/MT.gif) | *sprinting athlete, body in action, time-lapse* |
-| **EBA** · bodies | ![EBA](assets/gallery/EBA.gif) | *two figures, legs and feet, three people* |
-| **pSTS** · social motion | ![pSTS](assets/gallery/pSTS.gif) | *hugging, greeting handshake, three people* |
-| **V1** · early visual | ![V1](assets/gallery/V1.gif) | *pop art, circuitry PCB traces, monochromatic* |
+| Region | Synthesized stimulus |
+|--------|:--------------------:|
+| **FFA** · faces | ![FFA](assets/gallery/FFA.gif) |
+| **PPA** · places | ![PPA](assets/gallery/PPA.gif) |
+| **MT** · motion | ![MT](assets/gallery/MT.gif) |
+| **EBA** · bodies | ![EBA](assets/gallery/EBA.gif) |
+| **pSTS** · social motion | ![pSTS](assets/gallery/pSTS.gif) |
+| **V1** · early visual | ![V1](assets/gallery/V1.gif) |
 
 Explore the full interactive gallery and 3D brain maps at **[nevo-project.epfl.ch](https://nevo-project.epfl.ch/)**.
 
 ## How it works
 
-```
-STAGE 1 — Genetic search over prompts (image)
-════════════════════════════════════════════════════
-
-target (voxels / fMRI vector)
-        │
-        ▼
- image prompt gene pool ──crossover──► child genes ──mutation──► conditioning
-        ▲                                                             │
-        │                                                             ▼
-        │                                                    SDXL-Turbo (text→image)
-        │                                                             │
-        │                                                             ▼
-        │                                                           image
-        │                                                             │
-        │                                                             ▼
-        └───────────────── activation reward ◄── vJEPA-2 fMRI encoder
-        
-        (repeat over generations)
-                        │
-                        ▼
-        best image  (highest predicted activation)
-
-
-STAGE 2 — Genetic search over prompts (video, fixed image conditioning)
-════════════════════════════════════════════════════════════════════════
-
-best image ─────────────► fixed image conditioning (constant across search)
-                                        │
-target (voxels / fMRI vector)          │
-        │                              │
-        ▼                              ▼
- video prompt gene pool ──crossover──► child genes ──mutation──► conditioning
-        ▲                                                             │
-        │                                                             ▼
-        │                                              LTX-Video (image→video, fixed image)
-        │                                                             │
-        │                                                             ▼
-        │                                                        2-second video
-        │                                                             │
-        │                                                             ▼
-        └───────────────── activation reward ◄── vJEPA-2 fMRI encoder
-
-        (repeat over generations)
-                        │
-                        ▼
-        ranked candidate videos  (best_prompt, best_score, best video)
-```
+NEvo runs an **evolutionary (genetic) search** over prompts, guided by a brain encoder:
+
+`prompt population → SDXL-Turbo (image) → LTX-Video (video) → vJEPA-2 fMRI encoder → activation score → keep best → recombine → repeat`
 
-The search maximizes an **objective over the encoder's predicted fMRI response** (default `indices_mean`: the mean predicted response across the target voxels).
+Each generation turns candidate prompts into short videos and scores how strongly the **target region** is predicted to respond; the highest-scoring prompts survive and are mutated/recombined into the next generation. Over generations the search converges on stimuli that maximally drive the target voxels (default objective `indices_mean`: the mean predicted response across them). It runs in two stages — first evolving the image prompt, then the motion prompt on top of the best image.
 
 ## Installation