[quadrature]

Why are transformers a better fit for frame generation. Is it because they can better utilize context from the previous history of frames ?