Stable Cascade is currently the newest released diffusion model from StabilityAI. It has 3.6b parameters. That makes it bigger than SDXL. The unusually high latent compression makes this model special. The compression is 32x but goes up to 42x. (normal models have 8x compression) This high latent compression makes the model learn much faster. I think this gives the model a lot of potential. It would be nice if you could train an adapter for it. :)
Stable Cascade is currently the newest released diffusion model from StabilityAI. It has 3.6b parameters. That makes it bigger than SDXL. The unusually high latent compression makes this model special. The compression is 32x but goes up to 42x. (normal models have 8x compression) This high latent compression makes the model learn much faster. I think this gives the model a lot of potential. It would be nice if you could train an adapter for it. :)