Then, the brain encoder learns to align the MEG signals to those image embeddings. Finally, the image decoder creates a plausible image based on those brain representations.