merge tool: https://github.com/ljleb/comfy-mecha

merge recipe:

1.

model "291H_v10.safetensors" model_config="sdxl-sgm" merge_space="weight"

model "3dStock3dAnimeStyle_v20.safetensors" model_config="sdxl-sgm" merge_space="weight"

merge "weighted_sum" &0 &1 0.5

model "animeScreenshotMerge_v40.safetensors" model_config="sdxl-sgm" merge_space="weight"

merge "weighted_sum" &0 &3 0.5

merge "weighted_sum" &2 &4 0.5

2.

model "ΣΙΗ-1.5.safetensors" model_config="sdxl-sgm" merge_space="weight"

model "3dStock3dAnimeStyle_v20.safetensors" model_config="sdxl-sgm" merge_space="weight"

merge "weighted_sum" &0 &1 0.5

model "animeScreenshotMerge_v40.safetensors" model_config="sdxl-sgm" merge_space="weight"

merge "weighted_sum" &0 &3 0.5

merge "weighted_sum" &2 &4 0.5

3.

model "mechanimePart1.safetensors" model_config="sdxl-sgm" merge_space="weight"

model "mechanimePart2.safetensors" model_config="sdxl-sgm" merge_space="weight"

merge "tensor_sum" &0 &1 0.5 0.0

links to models used (v-prediction):

https://civitai.com/models/1773389/3d-stock-3d-anime-style-checkpoint

https://civitai.com/models/1298771/anime-screenshot-merge-noobai

https://civitai.com/models/1217??5/sih

https://civitai.com/models/1301670/291h

recommended steps/cfg: 24/1.1

recommended sampler/scheduler: euler cfg++ & ays+ (https://github.com/pamparamm/ComfyUI-ppm) (this node pack also includes negpip which is good for some negatives like duplicate/clone/4koma in u can put in positive instead of negative.)

recommended mod: https://github.com/Extraltodeus/Vector_Sculptor_ComfyUI (improves composition by adding nearby tokens)

-> POSITIVE prompt set to backwards at ~3.0

-> NEGATIVE prompt set to maximum_absolute at ~3.0

for demonstration purposes example images have no upscaling/postprocessing or any loras, only an artifact reduction pass with https://openmodeldb.info/models/1x-Manhwa-Null. workflow is in the image just drag it into comfy.

if i upscale them i usually do low denoise pass (0.25) with mixture of diffusers: https://github.com/shiimizu/ComfyUI-TiledDiffusion

example images use prompts i found around the site ^^

tldr just my favorite merge after some attempts