PSA (04-11-2025):
Hi guys, it seems suddenly the model is starting to become popular, i'm so glad you guys like it, it actually means quite a lot to me ❤️❤️
It's been some time since I've played around with generating images (I bought a 3d printer so im now printing my waifus and painting them :3) but if you guys got any critique I would still love to hear it, you can either just post a message here on the model, send me a pm or catch me on Discord (the name is Hapse there as well ^^) ❤️❤️
------
Flat Justice is a 2d/vector trace (danbooru wiki) anime checkpoint based on noobai v-pred checkpoints and some nice loras.
After using Hexbloom - Ionsyx Revised by Arctenox I started enjoying the 2d style and since I've been generating images for a long time but never made any checkpoints or lora's I figured why not give it a try and make my own.
Recommended settings
Positive prompt:
masterpiece, best quality, very aesthetic, absurdres, newest, Negative prompt:
worst quality, low quality, bad quality, bad, watermark, bad hands, extra digits, artistic error, old, worst aesthetic, loli, censored, halo,Sampler/Scheduler:
DPM++ 2M SGM Uniform
or
Euler a Beta
Steps: 30 (40 is even better)
CFG: 4-5~Versions - resources used
Version 1.2
CLIP:
Loras:
Vae:
Version 1.1
Loras:
Version 1.0
Checkpoints:
Loras:
Vae:
Description
Loras used:
Cryosibilum Style by Comte
Changed the CLIP to Yesmix XL (NAI-VPRED)
Added a bit of Illustrious XL 2.0 to the merge
FAQ
Comments (12)
Hope you don't mind my blowing your gallery page up. This model has a very clean look that adapts exceptionally well to any lora I throw at it.
Not at all!! im really enjoying your images :D you also inspired me to do a small test of all the different dmd loras https://civitai.com/images/96617687
Thank you for the praise and also if you have any critic i would love to hear it! ^^
Also, its nice youre testing loras because it might give me some ideas for the next version ;)
@Hapse Only downside isn't the model's fault. Since it's vpred, all my lora are epred, so the lighting is affected. Sometimes the lora doesn't get applied much at all and I need to actually bump up both the weight of the lora and the trigger words to see an effect (see the cubism posts).
Again, not the model's fault, but something to keep in mind when trying out the lora. When it works, it works very well because this model isn't imposing a harsh style.
@Hapse I also need to add extra weight to skin colors: "(dark skin color:1.5)" otherwise mostly ignored for non-white skin.
@clueless_engineer have you tried "dark-skinned female" instead?
and yea i might have to make an EPS version as well, i have tinkered with it a bit but its so hard to get the colors right >.<
WE NEED AN INPAINTING VERSION! PLEASE! This is the best model i've ever used.
how do you make an inpainting version?
I think maybe i am doing something wrong does this model works in forge also will it need a Vae.I am getting random noise on forge with the samplers u mentioned.
I dont use Forge but my bet is that its because its a vpred model and Forge maybe doesnt automatically detect it, did you remember to update Forge? I do know it works with reForge https://github.com/Panchovix/stable-diffusion-webui-reForge (and Comfyui, thats what i use)
Bring it back!!! Such an amazing model.
Great model It's great at most Holo but in swapping to 1.3 it seems to have lost some of IRyS' info so I ended up having to revert back to 1.2 whenever having anything with her in it.