iCoMix - Comic style Mix!
Thank you for all Reviews, Great Model/Lora Creator, and Prompt Crafter!!!
See on Huggingface iCoMix
Free Generate iCoMix V4 & V5 in Pixai.art
You can try with Comic "Artist Name", See at:
Similar but maybe you can find some interesting results :)
VAE : Blessed2 VAE from NoCrypt
Default Negative Prompt i use (u can add or remove depends on your need) :
(worst quality:1.4), (low quality:1.4), malformed hands, bad hands, mutated fingers, extra limbs, blurry, bad anatomy, bad proportions, gross, disfigured, jpeg artifacts, text, child, depth of fieldYou can try with Minimal Negative first:
(worst quality:1.4), (low quality:1.4)And u can add Negative Embedding if needed.
Thank you & Enjoy! Buy me a coffee
Description
Added:
MODEL :
Mtompunk
PikasAnimatedMix
ZombiMix V8
Vintedois
SamDoesSexyBlend
LORA :
sxzHotBloom LORA
sxzHorny LORA
sxzVividDarkness LORA
duskfallsIllustration LORA
hadesStyle LORA
All sample size is 640x1024, with no Hires Fix
Thanks & Enjoy!
FAQ
Comments (83)
iCoMix V04 Out Now! Thanks & Enjoy :)
Awesome! Can't wait to try it out!
@datmuttdoe cool! can't wait to see the results :)
what's new this time?
@bullseyetroll in this version we got more thick outline, and i think better anatomy, but pls try it first :)
Hello lostdog! your checkpoints are awesome! just a question, I would like to get a certain LORA trained or maybe a checkpoint and you do it so good, do you have a page where I can message you and pay you for something like this?
@Cube3D wow, thanks for ur appriciate!!! well i just want to starting make 'buy me a coffee' while im do something ;)
@lostdog that sounds like a great idea! let me know when you have the "send me a cofee" link and I'll be right there!
@Cube3D you can buy me a cup of coffee now :)
@lostdog Done!! and let you a message there!
@Cube3D cool thanks mate!!
I'm going through each LORA and writing down all the possible trigger words, but would it be possible to get a list of keywords that work and don't work? Or how exactly to call each LORA's results?
For example for the Jim Lee Style LORA, do we use the prompt "JimLeeDCComics" or just "Jim Lee"?
please make convert it to diffusers so we can train our faces with it
can u tell me how to do it?
hi can u pls try, is it already diffuser or not? im convert it from https://huggingface.co/spaces/anzorq/sd-to-diffusers
@lostdog i think this converter needs a ckpt file. but ShivamShirarao colab could do this I think .
1st - download everything from the hugging face and put all of it into a folder in your google drive.
2nd - check install requirements, login into huggingface, install xformers
3rd- SKIP BOTH “Settings and run” and “start training”.
4th- “specify the weights”. On the left hand side, theirs a folder button. Click it. Then it’ll open up a tab on the left with some options. Click on stuff till you find “Gdrive” or “My Drive”. Then click until you find that folder you put all the hugging face stuff in.
When you find that folder, click the 3dots on the right side and “copy path” and paste it onto the “specify the weights” thing. Now the computer knows what you want to work with.
5th- click convert weights and done
Since v03 I can't download, it just me?
The following message always appears:
This site can’t be reached
civitai-delivery-worker-prod-2023-04-01.5ac0637cfd0766c97916cefa3764fbdf.r2.cloudflarestorage.com’s DNS address could not be found. Diagnosing the problem.
yup its sometimes like that, but you can try it again, i think the site error
Awesome model! Any chance of an inpainting variant?
After some research I've realized anyone can make an inpainting variant of any model, if anyone needs an inpainting variant just look it up, it's actually super easy.
HOW THE HELL DO I SHOW MY APPRECIATION TO YOU SIR?!?
This is what I've done with your creation: https://twitter.com/J0HN9R1M3
wow awesome animation!!! well if u want to donate me, im sorry my ko-fi account was suspended coz some adult image, im trying to create again soon and put it here on my profile, but thanks and i feel great if someone really enjoy it with this model
@lostdog Thanks Brother, once your Ko-Fi is back up, let me know! Thank you again!
So disappointed I could get this model to do nothing. I get extra phantom characters deformities etc and I've tried all morning. Looks amazing though.
hmm whats wrong, maybe u can share the results
@lostdog I'll try and do that when i get a chance maybe someone can help.
ja acabo jessica? bagulho super facil tu ai sofrendo com oque mano ? se tu nao ta conseguindo nada é pq o problema é tu papo reto da pra gerar imagens fodas com literalmente 1 tag positiva 2 negativas
If you're getting phantom characters try (solo), if you're getting extra limbs try adding "deformed, disfigured, zombie, mutated" to your negatives.
Yea it could be your prompting, a lot of model mixes still have issues no matter WHAT we do to stabilize them XD
The VAE worked well for me for awhile.. now it just... doesn't. I get errors when I try to use it. it's disappointing. That said, when it works, it's amazing.
hmm, u can try change to none first, and load again
Same here but I think it's an issue with the VAE, not with the model. I'm getting NaN errors constantly now.
iCoMix V05 Outnow! :)
god bless you! that's my favorite model!
Lets go!
okay got some candidate, here some result https://www.deviantart.com/lstdg/art/Rogue-Thief-Girl-964082055, https://www.deviantart.com/lstdg/art/Superman-Ai-fanart-iCoMix-964081848, will put it more results ;)
Thank you for yoyur hard work
@fo3 my pleasure, hope can make great results :)
Should definitely put it on sinkin so it could be called through API
Unfortunately, v05 produces very pale/low contrast images on my system (AMD/DirectML, same issue for many manga style models). v03 and v04 worked fine, though, so I'll stay with the older models.
you use the blessed2 vae? i never try on AMD, but thanks for try it and the info...hope can fix it later...
Exactly the same for me
@elesmera_synovis okay maybe this model not good on amd, well i can't try that too coz not use it...
@lostdog Don't worry! I just wanted to report. I will stick with the older versions 04 & 03. They are amazing!
@elesmera_synovis sure, thanks for the info :)
Same problem here. All pictures get a pale / greyish tone on V5. Intel-User here, same goes for a friend of mine, we are using InvokeAI
@Ritter_Basgrimm u use the Blessed2 VAE?
@lostdog I have the same issue, whats MSE vae? I cant find it online (im on Nvidia)
@aweinspiring2017206 oopsss sorry not MSE for icomix use Blessed2 VAE
@aweinspiring2017206 here the Blessed2 VAE
sorry guys i mean Blessed2 VAE (i was thingking this AmIReal page lol)
Just in case it helps anyone else. I had the exact same problem and using blessed2 vae fixed it.
poor AMD users...
Thank you for this post. Made a world of diff. Was wayyy to pale. excellent post. Cheers
Not really working that well for me - getting bad and blurry faces, though the art style and prompts do create decent outlines. Shame it can't make faces that well for me though
maybe you can put the result so i can see, and also check the VAE you choose
@lostdog Yeah sorry my bad, was using the grapefruit vae instead of the blessed2 vae
@sidv cool, hope you can get great results now :)
Adetailer helps sometimes. Blesssed is a good Vae i still use WD tho lol
My dumb ass:
Hmm faces look familiar
Also me :
Oh yea there's lioke only about 10 comics models on the market right now adn wer'e all merging lora and back mixing each other's models.
Also me:
CACKLES ~_~ even if pooda isn't in this it' sstill amazing runs off like a child XD
yup, i merge and mixing again with old version model , and i think we have a few good comic model now, it will help me to make an update version, and hope i can mix it with pooda in next version XD, thank you!
@lostdog MAYBE BY NEXT VERSION i'll have not just POODA but more illustration models! My goal is to continually improve several lines -- XD I can't just stick with just one, not in my nature -- I have dissociative dentity disorder AND ADHD lmao -- so i'm like a rabid squirrel with different aspects that just confuses people like MAAAAAD lol - So right now i've got Pooda, Liquid Illustration, i'm seperating Osenayan's illustration and working on that soon - I'm giving up on the realism chase to focus on whati 'm good at :3
@duskfallcrew great! maybe i can use it, if you give permission to merge it :p, for now i interest with pooda illustration, i think more comic than the pooda comic. yup realism not more good than reality lol, i make Amireal maybe about 20-30 merge & delete, mix again, merge again, until i really find the good not perfect, and now i still try hard to make easy between man & woman, both genital & futa, (no hardcore) but keep the beauty :p
@lostdog You have full permission to even use any of my anime mixes XD - We're all in this together - I grew up on comics more than I did anime :3
@duskfallcrew cool! will turn that anime mixes as comics XD
@lostdog Exactly, and if you're on any social media or discord hit me up i got unpublished mixes not on civit yet XD
@duskfallcrew yup just saw you on DA, will find you at discord :)
@lostdog mm was that me> *kidding
@duskfallcrew Do you ALL have ADHD, or just some members of your system? 'Cause I feel ALL would be a terrible burden, but some or even most could be kinda useful.
@Alyndiar ;_; lol how dare <3 how dare <3 And we're not sure, but like on social media we try and lessen the odd but we're just all over teh damn place XD
Sad, but the same face for every girl..
I'm having this issue, no matter what I generate, all the images come out with this brown/grey filter. It is a shame because I really like what is coming out.
Example: https://imgur.com/a/c15X62t
Anyone know how to fix it?
try use Blessed2 VAE
did it work? @softpressureh260
@jetpackssk402 Hello! the same question
I really like the art style however is seems to freak out when trying to use tiling
Hi, just moved to v5 and running into an issue with the blessed vae - it keeps spitting out AttributeError: 'NoneType' object has no attribute 'pop'
any idea how to resolve?
Could you pleasssssssssssssssssse make an XL version of this? It's so damn good
This model would be so cool if made for either XL or even merged with FLUX, think about it!
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.