This model was trained of very close up photos of images on a cathode-ray television set. It doesn't exactly replicate the red, green, and blue bars, and the images can be a bit abstract or obscured, but it's interesting.