Simple animated merge that I use.
Positive: Masterpiece, best quality, amazing quality, very aesthetic, high resolution, ultra-detailed, absurdres,
Negative: lowres, bad quality, worst quality, bad anatomy, sketch, jpeg artifacts, ugly, poorly drawn, blurry, watermark, bad eyes, bright pupils, white pupils, bad hands, bad face, greyscale,
Euler A
30 Steps
Adetailer
CFG 1~5 - Try lower settings
Hires: 4xUltrasharp
I use the positive and negative embeddings from Illus Scores. https://civarchive.com/models/1062439/illus-scores
Post your images I like seeing what everyone comes up with.
Description
FAQ
Comments (9)
Great Model. I'm trying to use my LoRA's with it (more than 350 up to date) but your model seems to ignore them. Any idea? Thanks!
From what I can tell it looks like your models are for SD1.5 and this checkpoint is an Illustrious model, I'm guessing its just not compatible.
Where have you posted your images?
@Pinnacle You are right. All my models are LoRA SD1.5. Any similar checkpoint for SD1.5?. Results of this model are stunning.
@Pinnacle Do you know any way to re-create my LoRAs for an IL model like this one? Thanks!
@barabasj214 From my usage experience, SD1.5 Lora will not work with SDXL, (illustrious & pony are based in SDXL). Your loras generate photo realistic images of real people, photo-realistic loras don't tend to go well with animated models. However animated lora with realistic model can be hit and miss. To use your loras you need to use an SD 1.5 model.
I have created 0 Lora's, I am not the expert. I have never seen anything where someone was able to change a model from SD1.5 to SDXL or to FLUX, or to Aura Flow .... Etc. I believe that you will have to retrain models.
I prefer realistic models to anime my favorite SD 1.5 anime model is: AniMics - v3.0 | Stable Diffusion Checkpoint | Civitai
@barabasj214 I wish I knew but I've never made a lora before so I wouldn't know where to begin.
@Pinnacle I have bad experience training LoRAs for XL & Flux... that's the reason I still doing SD 1.5 LoRAs. I run trainings on my Local PC and the results are very good. For illustrations I use SexyToon3D models and RealisticVision models for realistic results. Thanks for your comments anyway.
Nice model. If you want a bit more realism, use cyberfix - subtract cyberrealistix xl from sd_xl_base and add it to this at 1.0. It also works ok with DMD2 strength 0.1 to 0.5 boosting speed and realism. With DMD2 lora at 0.5 it will generate nice images in 4 steps using cfg 1 (which ignores negative prompt) to 1.5 (which uses negative) using DPM++ SDE.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.