Studio Artist processed story driven generative image synthesis animation using stable diffusion. These are 10X time expanded (flow) with a small amount of enhancement processing.
Above is using recursive feedback of the previous frame to feed the latent representation of the U-net. Below interpolates the text prompt embedding with no previous output frame feedback.
I feel like the style prompt is dominating the storyline to some extent in these 2, so i've got some more experiments for a different remix post where we dial down the style strength.
No comments:
Post a Comment