Sign In

Soviet Brutalism SDXL V1.0 - v2.0 Showcase

06414-1169581155-CCCP,foggy,isolated,ramp,cantilever,movie camera,warm interior light,rusty,weathered,skyscraper,thick walls,partition walls,snow.png
06509-2922568120-CCCP,(foggy_1.7),ramp,movie camera,warm interior light,wildness,rusty,weathered,skyscraper,thick walls,partition walls,snowy,bar.png
06884-744962783-CCCP,_lora_Brutalism-000012_0.6_lbw=XYZ_,snowy,bare tree,dusk,foggy,symmetrical composition,exposed structure,foggy,beams,soviet.png
06906-66815145-CCCP,brutalism,snowy,bare tree,dusk,foggy,exposed structure,symmetrical composition,_lora_Brutalism-000012_0.6_lbw=test_,foggy,b.png
07092-600476336-CCCP,(soviet_0.9),(partition walls_0.8),(exposed structure_0.7),_lora_Soviet-000012_0.6_lbw=OUTD_,(inclined support_0.5),(snowy_.png
07165-1488961407-CCCP,(soviet_0.9),(partition walls_0.8),(exposed structure_0.7),_lora_Soviet-000012_0.6_lbw=TEST_,(inclined support_0.5),(snowy_.png

This version hasn't been fully completed yet due to the issues of WebUI's support for SDXL

What's New?

1.196 pics training dataset.(compared to previous 70 pics)

2.SD highres fix for all pics, and more aesthetic content.

BUT...

I encountered 2 major problems, and I hope someone can help me.

  1. The plug-in Supermerger for WebUI couldn't correctly merge Lora with block weight. For ex:

-when I tried to merge SDXL Lora with another SDXL Lora, in the format like this

CCCP-012 SDXL V2.0:0.6:TEST,Soviet-010:0.1

and I get error:

  1. The image2image function couldn't work with tile diffusion, which means I couldn't make image into Highres.(work perfectly with SD1.5 Model, but not for SDXL model) For ex.

    And I got error:


Test Process:

Recommended Lora weight <0.7

So I have tested it with Lora Block Weight, unfortunately I couldn't merge correct weight to this Lora

The Y-axis information is the Lora strength which I want to mix. (Also failed)

”TEST“ block weight compared to the original. (So what you expect now is the quality represented by “ALL”)


Really Hope Someone can help me with the SDXL Lora problem,

And I will quit WebUI and turn into ComfyUI after I finish this Lora.

if you like my Lora, welcome to create picture feedbacks,

Thank you.

Comments