Neural Network Predicts Brain EEG Responses to Text and Images

Researchers at the University of Florida published the Concept2Brain model in Nature Communications — a deep neural network combining CLIP and cVAE that generates synthetic EEG signals from text or images.

Compiled by Sergey KostenchukPublished 2026-11-14Updated 2026-11-14

2026-11-14 Research OpenAI

🤖 Neural Network Predicts Brain EEG Responses to Text and Images

The University of Florida published the Concept2Brain model in Nature Communications: it generates synthetic EEG signals from text or images. It utilizes OpenAI's CLIP and cVAE internally; training involved 13,640 EEG recordings from 88 participants, reproducing P100, N200, P300, and LPP components with a correlation above 0.6.

🌍 BCI teams and neurobiologists can obtain EEG datasets without human experiments and test hypotheses about the neural coding of emotions and semantics more quickly. The model accounts for individual differences (88 "virtual participants").

👤 Through an open web service, users can generate an EEG response to any concept or image without a laboratory and compare it with real data. Text-based predictions should be handled with caution: the model was trained on responses to images.

Source 1: https://www.nature.com/articles/s41467-026-75653-x Source 2: https://concept2brain.brainbodybehavior.science/

Sources