If you ran into any troubles with my mix, please read description below. There is plenty of useful information. Also, you can contact me through message system on CivitAI.
Если вы столкнулись с каким-либо проблемами при использовании моего микса, пожалуйста, прочитайте описание ниже. Там есть много полезной информации. Так же вы можете связаться со мной через чат на CivitAI.
Also try: DivineEleganceMix!
Introduction:
DivineAnimeMix, obviously, more anime-styled version of DivineEleganceMix, that was inspired by more clean and straight-forward anime style. It's still keeps some details of original mix, but also has it's own more unique and anime-ish visuals.
Clean and sharp anime style
Less post-processing in generations
Great prompt guidance
Great character and concept recognition level
Great LoRA (nai/any-based) compatibility
I hope you also will like DivineAnime, so stay tuned for newer versions of mix!
Recommendations:
Use danbooru tagging for prompts
Sampler: "Euler a" at 30 steps
Cfg scale: 8
Clip Skip: 2 <---- VERY IMPORTANT
ENSD (Eta noise seed delta) - leave blank
Recommendations (Hires Fix):
Upscaler: Fatal Anime
Upscaler: R-ESRGAN 4x+ Anime6B (basic)
Hires steps: 10 (Euler a)
Hires steps: 15 (other samplers)
Denoising strength: 0.3 (Euler a)
Denoising strength: 0.4 (other samplers)
Upscale by: any
Troubleshoot:
My colored are bleached out. What should i do?
You're missing a VAE file. For this mix i would recommend kl-f8-anime2 VAE. Just put it into SD folder -> models -> VAE folder. Then go to your WebUI, Settings -> Stable Diffusion on the left list -> SD VAE, choose your downloaded VAE.
If you want to get mostly the same results, you definitely will need negative embedding:
EasyNegative, it's better to use it at 0.8 weight.
Don't use EasyNegative (or other embeddings) with Style LoRAs, it may ruin the style!
Until you like the result, of course.
Don't use a lot negative embeddings at the same time. It definitely will ruin any possible output on all models! If you want to use negative embeddings - use one for quality, and one for hands (if you want a placebo for hands).
Technical support:
If you're ran into any problem with my mix, you can DM me on discord:
troubledarkness - discord nickname.
I'll be glad to help | Буду рад помочь.
Credits:
Really BIG thanks to:
andite for Cocoa
Linaqruf for original Anything model
You, for reading, downloading and reviewing this mix
Description
It was a long time since first version release, so, lets start.
Version changes:
Global:
Changed merging weights
Removed Cocoa from the mix
Now models has more clean anime look
Characters:
Bigger, better and more deep eyes
Smaller torsos in overall
Easier-to-prompt face expressions
More reactive bodyshape prompts "curvy", "skinny" etc.
Environment:
More control over fog and blur
Now fog can be removed with "fog" and "mist" tags in negatives
More blur for backgrounds, can be turned off with "blurry" in negatives
With this version of DivineAnime all your LoRAs should work as good as on previous version, so, there is no any problems with compatibility.
But can do bigger eyes on some style LoRAs
Thanks for keeping an eye on Divine series models and thanks for reading this changes, i hope you'll like new version of DivineAnimeMix as well.
FAQ
Comments (18)
It was a long time since first version release, so, lets start.
Version changes:
Global:
- Changed merging weights
- Removed Cocoa from the mix
- Now models has more clean anime look
Characters:
- Bigger, better and more deep eyes
- Smaller torsos in overall
- Easier-to-prompt face expressions
- More reactive bodyshape prompts "curvy", "skinny" etc.
Environment:
- More control over fog and blur
- Now fog can be removed with "fog" and "mist" tags in negatives
- More blur for backgrounds, can be turned off with "blurry" in negatives
With this version of DivineAnime all your LoRAs should work as good as on previous version, so, there is no any problems with compatibility.
- But can do bigger eyes on some style LoRAs
Thanks for keeping an eye on Divine series models and thanks for reading this changes, i hope you'll like new version of DivineAnimeMix as well.
Thanks for putting in all the work to make quality models!
I have a question for you reguarding v2 vs v1.
The samples you give for V1 look slightly more... 3d? shaded?
Was that due to your "removing Cocoa" ?
What if we would like all the improvements you made to v2, but also would like that original style?
@phil866 I want to improve my models further, of course. I've removed cocoa from mix because it was causing at least eye quality drop and if "lying, on back" was prompted, it was doing a lot of pictures with character lying upside down, i mean, with head near "camera", that also causing a massive face quality drop and anatomy issues. There was more issues i don't remember right now, but with block merge, i hope i'll be able to do a big quality update for DivineAnimeMix someday.
And yeah, v1 was more detailed and "realstic" or shaded, as you said, because of Cocoa in the mix.
WOW ! V2 ! You are the best ! Thanks for keeping this model updated
Thanks for commentary. I don't know if you tried the model, but for me it's a huge update for DivineAnimeMix and its visuals.
@TroubleDarkness I just tried it a little bit ... to be honest with you, I prefer the v1. The v2 changes the character (the head is a bit too thin in my taste), it adds a bit too much detail/sharpness. I will stick with the v1 for the moment. Still that's cool that v2 exists :)
Good model, but if "back" is in the prompt, it shows the character's back instead of "arms behind back".
Well, i've seen this kind of behavior on DivineEleganceMix, but never got anything like from behind shot on DivineAnimeMix, except "backlighting" prompt, which can really lead to from behind shot. Will try to find out the problem, if there is.
Hello, it looks very good! Does the training dataset have male characters too?
If you mean LoRA merged in, yeah, there was male characters.
The model is amazing, but I don't get two things..
It's a merged model, but you can't use it in a merge, also, the images you generate you can't sell or use it for commercial work.
Why so many limitations??
Hi,
about merges, i'm pretty sure i've never limited any merges sharing, maybe i forgot to check the mark after v2 update on this permission, not sure. Anyway, fixed. But talking of commercial use, it will stay limited.
Some day i'll enable on-site (CivitAI) generation, but i'm waiting for hires. fix option to be added.
@TroubleDarkness got it.
Thank you for clarifying. 🥰🥰🥰
Perfect model :)
Hi, thanks for commentary! I'm glad you've liked this model. Maybe some day i'll be able to update it to bring even better anime style and restore more concepts as possible after merging.
This model is a great addition to the Divine Elegance Mix! Having the same good qualities, it is more cartoonish and colorful, and, as it seemed to me, more flexible and responsive. Great job! Thanks!
Some people was arguing about v1 of DivineAnimeMix be better at visuals than v2, you can also try it, if you didn't. Second version is more "cartoonish" and there a lack of visual details, in my opinion. It was a lot harder to balance this mix between simplicity and visuals. But i guess, it will not get any further updates.
Important (not so) note.
For those, who wonders (maybe, for some reason), CivitAI just don't let you run DivineEleganceMix and DivineAnimeMix (and a lot of models, actually) on their generation service because? No idea. Even the newest models (SDXL, Pony, FLUX) uploaded even few hours ago by new users may have this option available from start. Both models (Elegance / Anime) has "CivitAI generation" turned on in the settings for nearly half of the year, but CIvitAI just didn't opt-in this model for generation service. No idea how this works on their side. Most models with the same ratings / downloads / upload time, etc are available, and mine are not. What a great day to be an author on CivitAI.
Thanks for your attention. Just wanted to let you know about this situation. I hope you had great time using Divine series models in the past, or maybe even having it right now.
I'm planning to get back to LoRA training in the near future. For PonyDiffusion-based models. SD 1.5 may have better visuals sometimes, better backgrounds and some knowledge in simple concepts like tree types, etc. But it's a dead end. I hope you'll find those releases cool or at least, useful. Hope to see you soon.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.