bvga Styles :: Models I use to generate my images

Update 2:

After experimenting with my style LoRA, I've come to like these style variants:

Checkpoint: Wai Illustrious V17
I really like the colors, details, and style this gives. It's a step up from v16, which had washed out colors and foggy backgrounds.

Core Style:

0.7-0.9 strength: Core bvga v2.0

This gives a good anime-feeling style.

ATNR Variant:

0.5 strength: Core bvga v2.0
0.5 strength: ATNR

This gives a more detailed style reminiscent of the old 8-lora stack.

Hyperrealistic Addition:

0.2 strength: Hyperrealistic

Adding hyperrealistic on top of either of these styles gives it a semi-realistic feel.
(However, it may also result in lighting and color issues, which you will need negative prompting to address. I found this one generally works: "harsh shadows, blown highlights, overexposed highlights, underexposed shadows, extreme contrast, color cast, strong color tint")

________

Update 1:

You can also try out the style LoRA I trained to replicate this style, all with a single LoRA!
https://civitai.com/models/2433005/core-bvga-style
It doesn't work entirely the same, as I trained it on my own images (it's not a merge of the models). But it captures what I was trying to accomplish with the style.

v2.0 of my style LoRA is what I generally prefer to use now, as it gives much better prompt understanding, creativity between seeds, and mixes well with character LoRAs.

________

Original Article:

I've seen some interest in the models (and settings) used to generate the images I use for my videos, so I thought I'd share them here for anyone interested!

(I've also attached an image containing a comfyui workflow with this information)

Sampler:

5 CFG

30 Steps

Dpm++ 2m karras

LoRA triggers:

(I place these at the end of the prompt)

atnr, novuschroma74 style, 7481llu, Hype4realistic, uh_sk1n

Checkpoint Model:

Wai Illustrious V16

https://civitai.com/models/827184/wai-illustrious-sdxl

The checkpoint isn't as important as the loras used; the loras are what define much of the particular style, and can be used across other anime-styled checkpoints.