Regards for all your love and support ❤
avoiding negatives can improve result on SDXL
Description
Trying to get the best of both versions V5 and V6
FAQ
Comments (40)
The HANDS are fantastic for V7.
Thank you Op!
V7 is amazing! Thank you!
Hi.How does it work with Hyper/Hyper-CFG-Lora?
Is there a chance for a version with turbo blended in? Those are SDXL models which work the best with my limited hardware, still producing nice results. I'm a big fan of your SD 1.5 stuff.
The Skin Texture Is Good!
some ideas to improve model:
- augmentation with image slicing. you can slice image in parts where model is poor to generate, for instance, if model is draw poor foots, so the same image of a full body woman, you may cut foots and use any upscaler, then use it as input to train in complement with original image.
- full body images tend to be deformed, I guess it because of images used on dataset.
- I could not be able to generate realistic people without loras, I guess its related to the dataset origin
- perhaps you could share a list of key words used to finetune the model to have a cheatsheet, it will help a lot.
If I remember anything can help I'll send here again.
Thank you so much for finetuning this model 👏
It always helps to share images if u make these statements. Probably work on your prompting if u are not able to produce realistic outcomes without loras.
Thanks anyways :)
v8 when?
Why? It is still the best XL model out there 😅
@epinikion How about a turbo model
@devsinghushan293 Is there really a use for that?
@epinikion There are no good turbo or lighting models, if you make a turbo model it will be on top of the list just like your epicrealism sd1.5 models.
@devsinghushan293 did u test my Lightning model? Tubo is crap due to its trained resulution 512, 768
@epinikion I tried it but it is giving me humans with long torso and fused humans, any solutions? I used the zues version btw with 1216x832
@devsinghushan293 Pls try out the new Hades Version https://civitai.com/models/354130?modelVersionId=534487
@epinikion You are a legend dude! AWESOME CHECKPOINT
@devsinghushan293 Thank you
May I ask how do you guys loading this fine-tunes safetensors file with diffusers? I cannot load it into base sdxl1.0 model pipepine :(
This V7 model is insanely good. Of the ones I've tried this seems to be the best for SDXL. A close second would be Albedo base.
It's not just the quality, it's the variety. Many other checkpoints seem to be identical when it comes to strange prompts, but this just has more flexibility.
Very diverse model - seems to work great in a lot of situations
-I wouldn't say it is better in comparsion to LEOSAMs Model, so i would recommend you to use his model instead of mine.
Props for saying that ^^ thanks for all your work
Thanks. It was meant for the V3 version but many guys and girls didn’t like the merge with hello world. So I was going back to refine on my own with new setup. Everyone has other likings. I like the V7 pretty much now. 😉
Any chance of a inpainting model?
So far so good. Still testing. It translates prompts from other models pretty nicely so far and gives realistic results.
Which version of LEOSAM's HelloWorld XL did you use? You do not specify which version.
I linked to the exact Version in the Version Notes
very nice results
Can you take this word back "probably last Version for SDXL until SD3" ?
epiCRealism is EPIC in terms of everything.
Thank you! I love your creative creations!
@epinikion Thank you for sharing this incredible model; I've been using the epic versions for over a year and have been a fan since.
I absolutely agree
I tried my best to get good prompts with Juggernaut but gave up. Been consistently using RealVis for months now. Just checked out this one, and it's a game changer. Thanks for all the work you put into this.
Thank you for you kindness.
Same here, I just can't get decent stuff with Juggernaut XL... Maybe it's my stryle of promptimg, who knows. But EpicRealism XL absolutely rocks. It has been my main realism model for the last 3 months.
I would love to see a more "serious" pony model
What do you mean by that?
More in the style of your epic models. I really like the style.
Ohh. Sorry. I just recognize that there is a brand new one.
@Pressydent I probably have another finetune ongoing, but I don’t get the danbooru tagging stuff … But yeah I have another attempt for it 😅
Details
Files
epicrealismXL_v7FinalDestination.safetensors
Mirrors
epicrealismXL_v7FinalDestination.safetensors
epicrealismXL_v7FinalDestination.safetensors
epicrealismXL_v7FinalDestination.safetensors
epicrealismXL_v7FinalDestination.safetensors
epcrlsm_v7.safetensors
epicrealismXL_v7FinalDestination.safetensors
epicrealismXL_v7FinalDestination.safetensors
epiCRealismXL.safetensors
epiCRealismXL.safetensors
epicrealismXL_v7FinalDestination.safetensors
epicrealismXL_v7FinalDestination.safetensors
epicrealismXL_v7FinalDestination.safetensors
epicrealismXL_v7FinalDestination.safetensors
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.