Trained by natural language. Better finger performance.
Using the latest non-monochrome image to train.
Version W is aiming the white style
Version Y is aiming the yellow and dark style
