I've been working on my own GAN for a while, trained on a few different datasets. Then the EEG data is fed into the GAN in real-time and used to interpolate within it's latent space. I assign each brainwave band (e.g. alpha, beta, etc.) to it's own point within that latent space, so the core idea is that as alpha increases (for example) the output heads towards that point in the space and that's reflected in the output image. Happy to talk more.