trained off of 90~ images plus frames from animations, use with a weighting of 1-2.