Type | |
Stats | 615 2,909 |
Reviews | (85) |
Published | Sep 1, 2024 |
Base Model | |
Trigger Words | graphic novel comic book style sketch lineart ligne claire |
Hash | AutoV2 C5D7E4579C |
IMPORTANT : Use this VAE instead.. The one took on Civitai list seems to be a problem : https://huggingface.co/madebyollin/sdxl-vae-fp16-fix/resolve/main/sdxl_vae.safetensors?download=true
Suggested to me for the time I spend on these models and if you like what I do here, please considerate to donate on this Ko-Fi page : https://ko-fi.com/aurety
__Version 1.8 update_______________________________________________________________________
Hi friends, it's been a while hasn't it?! I know you've got a lot of exploring to do with Flux things but maybe you can spare a little time for this update of CHEYENNE ( probably its last in SDXL version )... I've been working on it for a few months and the style is more speed painting and concept-oriented than comics, but I'm staying true to my desire to build production-oriented models and this one is incredible!
More coherent update and refinings. Thanks for support ! Hope you will enjoy it.
This version is extremely sensitive to the nature of the sampler, steps and CFG so you'll have to test what's best for you but overall :
- DPM++ 2M // 10-30 steps // CFG 1-5 will essentially do the job.
- DPM++ 2M SDE and DPM++ 3M SDE are excellent alternatives, with images that are often crisper and more accomplished (but less sticky to the prompt).
I hope you enjoy this, and hopefully the next OPUS will be a FLUX version! Kisses and take care!
Version 1.6 Update ___________________________________________________________
The update I'd been wanting to do for a long time: 3 weeks of hard work, over 300 mergers... At last I've arrived at the foundations of the model I'd originally envisaged. There's still room for improvement, but now I'm going to concentrate on the 'production' part and make it a more specific tool for people in the business, or for my students and friends.
Thanks to Duskfallcrew for his support on KO-FI, it means a lot to me to have the recognition of the people you love. Thanks also to Marc Quentel and Thomas from Stablediffusion.blog for your precious advices. This is probably Cheyenne's last public update, so make the most of it! Subsequent updates will be discreetly passed on to my followers.
Thank you for all the encouragement I've received and those who've posted in the gallery. You are beautiful and brave. I love you all!
>So what's new?
A massive update of styles and flexibility (this involves better management of the CFG, which now plays an essential role! From 1.5 (rougher, sketchy) to 4-5 (good medium mix) to 7-8 (more contrast, flatter, outline)... and also more realism (so comics, or graphic novels illustration are the basic keywords for using <em>Cheyenne</em>1.6)
- Able to go from a painting style to flat colour, with or without outline, all the usual keywords are present: painting, comics, outline, ink, drawing, etc...
More influence on the style of European comics: Hergé, Spirou, Franquin, Uderzo, Gibrat, Moebius, etc. are now recognisable. There's a bit of Manga, but very little: Ghibli, Junjo Ito, Makoto Shinkai, etc. )</p><p>I'll leave you to explore the rest! It seems to me that the model is more complex but also more extensive than the previous version. I hope you'll show me in the gallery...
My models are really production-oriented, and it's with that in mind that I'm improving them. If you want to help me to acquire a new GPU ( Or you have better than an old 2080Ti to offer me, LOL ! I'll take it !!! ) or if you like my work, I have a KO-FI open, or you can follow me on Civitai or on my X/Twitter: https://twitter.com/AAurety
Kisses !!
____________________________________________________________________Cheyenne v1.4 Update
First of all I'd like to warmly thank my 2 KO-FI donors: Ja and glacier434... Thank you for helping me to get a better material to continue this adventure. My models are designed for PRODUCTION and concept art. I have so many ideas but few resources... Thanks also to Scott Detweiler and Grokster for exposing MOHAWK in their media and nice words from them ! I love you guys !
Follow me on KO-FI or X : @AAurety and please considerate to donate if you can, or to follow my progress on CHEYENNE, MOHAWK or the new model for Matte painters SIOUX that I'm working on at the moment which is crazy !
In short, CHEYENNE V1.4 is a MASSIVE update! It's just incredible ! I'll let you discover the images in the gallery but I'm really happy with this update. I hope you like it as much as I do.
The concept remains the same: a European-style cartoon-comics model, with the aim of making comics, charadesign and concept-art for video games that is creative, highly graphic and flexible... This version satisfies me in every way...
Note: You may get more realistic renderings than expected but add lineart, graphic novels, or any other related term to change that. Side effect to change a bit the Manga Bias..
.CFG : 1.5 ( grungy, creative ) -> 4-6 BEST -> 8 ( vibrant, finish )
Sampler : DPM++ 2M Karras ( outline, classic ) // DPM++ 2M SDE Exponential ( more vibrant ! )
I also like to use different VAE depending on my needs ( see below )
https://civitai.com/models/152040?modelVersionId=170078
Kisses !
_________________________________________________________________________________________
V1.2 Update! It's a great update and I'm starting to get my wishes for this model. I wanted a template that wasn't too "biased" by the Manga or American Comics style, with the possibility of rendering visuals for European-style comics, but also storyboard, concept art, chara design, etc... With the choice of inking, line and style... It's not all sorted yet, but I'm heading in the right direction!
- A lot of work on creating comic strips and storyboards.
- Huge overhaul of the 'clear line' concept, now finer but clearer even at long distance.
- Adjustable inking via a few key words ( ink, graphics novel, sketch, etc.)
- Rich details and complex compositions
- Work on the prompt so that it's no longer necessary to write long texts.
It's not all sorted, there are still a few 'photos' coming out, but with the right words, they can be corrected... There are still a few problems with the anatomy, the hands, etc., but overall the faces are better, and so are the backgrounds!
I'm now using the VAE xlVAEC_c1.safetensors instead, which gives me a higher contrast image.
As a sampler, I use DPM++ 2M Karras or DPM++ 3M SDE Exponential, which give me good results.
Don't hesitate to post your images as they give me markers to improve this model or let me know your criticisms!
Love you!
___________________________________________________________________________________________
V1.0 update ! After more than 100 merges of LORAS and Modeles, I've arrived at this version, which I love and which exceeds my expectations. The aim of this project was to provide me with concepts, rather than finished images, so I made sure that LOW CFG ( between 1.5 and 3 ) gave me rich, detailed results ( see the "battle against the dragons" series in the gallery ) and I think it works pretty well. The renders are fast.
Prompts can be short and literal and the negative prompt is not always compulsory. The template remains flexible but it can "drop" photos rather than illustrations, so feel free to add words like: graphic novels illustration, sketch, comics, lineart, etc...
I still have a few biases on eyes, hands, and anatomy in general but frankly, I'm super happy with it and I hope you'll enjoy it too.
Take care and peace on earth!
________________________________________________________________________________
CHEYENNE V0.8 is an update to fix the most obvious flaws in version 0.5 and give it a bit of body and structure for its further development. So I've revised :
- refining the features and lineart in general (but I've still got some work to do on that!)
- the hands, eyes and general anatomy in part. It's not perfect but there's clearly room for improvement!
- low CFG work, because I've discovered that he doesn't do so badly in low CFG and low steps, as well as with the different VAEs... I strongly advise you to try the 3 magnificent VAEs from atw44qb ( https://civitai.com/user/atw44qb ) which I think work really well with my model. : https://civitai.com/models/152040/xlvaec
Thank you to all those who have posted their images and those who keep this community alive and thriving through their exchanges and sharing.
Kisses and Peace on the World!
_________________________________________________________________________________
v0.5 (BETA) :
CHEYENNE is the result of an error in the fusion planned for MOHAWK .
In the end, it was an obvious choice because I wanted to create a model for cartoons and comics, to push 2D a little further without altering MOHAWK too much.
For now, consider it as beta version !
I also wanted a model that was more representative of the Franco-Belgian comic strip style than of manga, and that was flexible enough to create character and landscape concepts for my classes... It's a first draft and far from finished, but I hope it will find a place in your projects.
May peace be with you and your loved ones, we're going through difficult times and it's thanks to these communities of sharing and exchange that there remains a bubble of hope and creativity....
Kisses to you.
I use DPM 2+ M Karras and ultrasharp as refiner but not needed all the time.
CFG 3 to 8 is ok.. for my early tests.. It's a beta version !!