beach synthesis

 

Generative animation constructed from a series of still images generated using CLIP image embeddings as input to a latent diffusion system (custom model weights for stable diffusion trained by lamba labs).  So no text embedding, just the CLIP image embedding from a painting of a beach scene.

For the second example below i used a real image taken at the beach on Maui's south shore.  That was fed into the CLIP model to generate an embedding that was used to steer the latent diffusion generative image synthesis (no text prompting in the system).