README.md

---
title: TRIBE V2 — Brain Response Prediction
emoji: 🧠
colorFrom: indigo
colorTo: yellow
sdk: gradio
sdk_version: "6.11.0"
python_version: "3.12"
app_file: app.py
pinned: false
license: cc-by-nc-4.0
hardware: cpu-basic
---

# TRIBE V2 — Brain Response Prediction

Predicts fMRI brain responses to **video, audio, and text** using Meta's TRIBE V2 foundation model.

## Features
- **Text Scorer** — Paste a script/hook, get brain engagement scores (~30s)
- **Video Scorer** — Upload a video for full multimodal analysis (~2-5 min)
- **A/B Tester** — Compare two text versions head-to-head
- **API** — Programmatic JSON access

## How It Works
TRIBE V2 combines LLaMA 3.2-3B (text), V-JEPA2 (video), and Wav2Vec-BERT (audio) to predict cortical surface activations across 20,484 brain vertices. Scores are derived from region-of-interest analysis using the Destrieux atlas.

## Scores
- **Attention Capture** — Will they stop scrolling?
- **Emotional Valence** — Does it trigger feelings?
- **Language Processing** — Is the message clear?
- **Visual Imagery** — Are visuals compelling?
- **Viral Potential** — Composite engagement score

## Citation
```
@article{tribe2024,
  title={A Foundation Model of Vision, Audition, and Language for In-Silico Neuroscience},
  author={Meta FAIR},
  year={2024}
}
```