 Speech 2 EEG is a novel approach to improving the accuracy of electroencephalogram, EEG, recognition. It uses pre-trained speech models to extract multi-channel temporal embeddings from EEG signals, then combines them into a single feature vector before feeding it into a classification network. This method outperforms other existing approaches on two challenging motor imagery datasets, demonstrating its potential for future research. This article was authored by Jean Zhao Zhou, Yi Chun Duan, Ying Ying Zou, and others.