Rockman AIさんのイラストまとめ

The Stable Diffusion 2 depth model has recently become widely available, and it dramatically improves image-to-image, making things like sketching a drawing and using SD to fill in details and shading much more feasible.

Sky Lagoon Remix V2, with depth awareness:

Trying to stylize Stable Diffusion 2.0 by training style embeddings for two Rockman related house styles: Teppen and X Dive. This uses both.
#megaman #ロックマン

@Slime_BlaXun Here are a few raw generations. None are perfect, but SD 1.5 generations would usually be missing at least one limb and would pose very stiffly.

Posted the previous attempt b/c it was interesting, but it didn't truly capture Giger's aesthetic. Second attempt at X in the style of Giger, using new "Aesthetic Gradients" technique trained on 10 Giger paintings. https://t.co/0qHY1T93pc

Zero. Custom Stable Diffusion model trained on official Zero artwork, with manual touchup and inpainting.

Energen Crystal Mine remixed with Stable Diffusion

Flame Stag's stage remixed with Stable Diffusion

Sky Lagoon remixed with Stable Diffusion.

Zero. Custom Stable Diffusion model trained on official Zero artwork, with manual touchup and inpainting.

Process:
Left: native 512x512 model output based on text prompt

Right: double resolution, manually resketch problem areas and inpaint (mask a selection from the image and force model to refine it while freezing the rest of the image).