V10 is sponsored significantly by Wizz (https://civarchive.com/user/_Wizz_) - check him out, amazingly kind and knowledgeable guy.
Please consider joining my Patreon so I can keep most of my work available for everyone - I'll be releasing early access models there for a cheaper monthly fee than buying them here individually, and also will be providing exclusive models and more!
For business inquiries, commercial licensing, custom models, and consultations, please get in touch at [email protected] or [email protected]. You can also contact me here through CivitAI DM or join my Discord.
If you see this model on a commercial product/website that is not CivitAI, then the website you are seeing it on is stealing and profiting on my work. Please contact me. If you currently offer it commercially, contact me to avoid being called out publicly.
Note: commercial licensing is a requirement for all generative services other than CivitAI. It's also a requirement for companies using it in a commercial setting; e.g. marketing materials, fashion shoots, apps for photo/video editing, etcetera. The only commercial usage that does not need a commercial license is for artists selling their imagery as prints and such.
Consider using DPM++ 3M SDE Exponential or DPM++ SDE Karras for this model, I personally like them best, 3M SDE Expo for most styles, DPM++ SDE Karras for photorealistic. Make sure to read the prompt advice further on in this section if you want to create photorealistic images as well as some of the sample images.
Introduction
A model line that should be a continuance of the ZavyMix SD1.5 model for SDXL. The primary focus is to get a similar feeling in style and uniqueness that model had, where it's good at merging magic with realism, really merging them together seamlessly. Of course with the evolution to SDXL this model should have better quality and coherance for a lot of things, including the eyes and teeth than the SD1.5 models. This model has no need to use the refiner for great results, in fact it usually is preferable to not use the refiner. Recommended to use ultimate SD upscaler to get the most amazing results.
Continue reading for more information and some tips and tricks.
I kindly request that you share your creations both here and on my Discord server, as I would greatly appreciate the opportunity to see them and motivate me to spend more time in further ventures here.
Pros and cons
Pros
Much better saturation than base model.
Much better teeth, eyes, hands and feet.
Increased realism.
Less blurry edges but still keeps a certain pleasing softness to the image.
Better looking people.
Great texture and tonality.
Cons
NSFW much better than base, but still somewhat lacking without LORAs.
Roadmap
Training the SDXL model continuously.
Pioneering uncharted LORA subjects (withholding specifics to prevent preemption).
Tips
To better understand the preferences of the model, individuals are encouraged to utilise the provided prompts as a foundation and then customise, modify, or expand upon them according to their desired objectives.
Ditch the refiner, an img2img ultimate SD upscaler gets better results when you select this model for it.
For photorealism, use nmkdSiaxCX_200k for initial upscale, consider using ultramix_balanced for final upscale/pass to lessen grainy pictures.
Consider using face restoration techniques when the result for the face is subpar, but the rest of the image is interesting.
If you find that the details in your work are lacking, consider using wowifier if you’re unable to fix it with prompt alone. wowifier or similar tools can enhance and enrich the level of detail, resulting in a more compelling output.
ComfyUI is the UI I use for my SDXL images.
Your 1.5 LORAs won't work in SDXL.
Consider finding new prompts, don't use the standard 1.5 ones. SDXL likes a combination of a natural sentence with some keywords added behind.
To maintain optimal results and avoid excessive duplication of subjects, limit the generated image size to a maximum of 1024x1024 pixels or 640x1536 (or vice versa). If you require higher resolutions, it is recommended to utilise the Hires fix, followed by the img2img upscale technique, with particular emphasis on the controlnet tile upscale method. This approach will help you achieve superior results when aiming for higher resolution outputs. However, as this workflow doesn't work with SDXL yet, you may want to use an SD1.5 model for the img2img step.
Prompts
Recommended positive prompts for specifically photorealism: 2000s vintage RAW photo, photorealistic, film grain, candid camera, color graded cinematic, eye catchlights, atmospheric lighting, macro shot, skin pores, imperfections, natural, shallow dof, or other photography related tokens.
Recommended negative prompts: As few negative prompts as you can, only use it when it does something you do not want, like watermarks. Consider using high contrast, oily skin, plastic skin if the skin is too contrasting or too oily/plastic. Also make sure to add anime to negative prompt if you want better photorealism, and more mature looking characters.
You are further encouraged to include additional specific details regarding the desired output. This should involve specifying the preferred style, camera angle, lighting techniques, poses, color schemes, and other relevant factors.
Recommended settings
sdxl_vae.safetensors (baked in).
DPM++ 3M SDE Exponential, DPM++ SDE Karras, DPM++ 2M SDE Karras, DPM++ 2M Karras, Euler A
Steps 20~40 (lower range for DPM, higher range for Euler).
Hires upscaler: nmkdSiaxCX_200k, UltraMix_Balanced.
Hires upscale: Whatever maximum your GPU is capable of, but preferably between 1.5x~2x.
CFG scale 4-10 (preferably somewhere around cfg 6-7)
Lightning LoRA specific settings:
Euler sampler with SGM Uniform as Scheduler.
Steps 4 (use the 4 steps LoRA)
CFG scale 1-2 (CFG 1 at the higher weights for the LoRA)
LoRA weight 0.6-1
Model recipe
Trained from SDXL 1.0.
Social media
You are welcome to join my recently created Discord server, where we can engage in discussions, share our experiences in AI, and showcase the things we’ve made with AI. You are encouraged to join and ask any questions or seek additional tips and tricks related to my models or AI in general. Your participation would be greatly appreciated.
Description
Trained extensively on a fine-art and fashion photography dataset. Still very much remains an all-in-one (AIO) model, but is now even more capable of realistic photography (with a lean towards professional shots).
Hair much softer and with more individual strands visible
Skin texture very lifelike
Full body and more diverging poses are easier to get
Quality of more distant people and objects is generally better
May diverge from some styles compared to v8 with the same prompt, may need more weight or less on critical tokens.
FAQ
Comments (32)
So according to Your prompts , it can go beyond 75 tokens?
Every SDXL model can, it just chops it up in blocks of 75. It's better to keep it within 75, so it doesn't chop up the concept's idea at a point where you do not want it, and also gives a quicker generation time. But it's all besides the point, you can go beyond 75. As for the sample prompts, I was lazy this time and just let OBP run through random prompts for a few hours and selected interesting ones to upscale. It isn't how I'd make the prompt myself.
@Zavy got it, thanks , i Will check The new version
@P_Universe or you can use BREAK to chop them yourself
Most prompts are full of non sense spam that don't highly influence the prompt after 75 characters anyway... I would avoid doing this, unless you like random stuff and high variance. 75-100 is the sweet spot for me most of the time.
For me the Zavy Chroma XL models are the best SDXL models so far.
v9 seems to be better for photorealism and anatomy. Unfortunately more logos. v8 is a little bit more creative - what I really like.
That's fair. Give the various painting styles you may want more weight and it should start to resemble V8 more closely for that. With the same prompt as V8 it'll lean more into photorealism though, yes.
@Zavy Many thanks for the great SDXL models from you. I will try to take your advice to heart and play with the weights a bit. It is a lot of fun to work with ZavyChromaXL.
If you want more creativity influence the generation more with creative loras... That is what most people use them for. To spice up the generations.
@Allure_Designs That is right, but might only help for a certain style, world morph or a concept/thing/person missing in the basemodel. By experimenting with promting we can get amazing results without using Loras and are much more flexible. I do not just want to "spice up" the result. If you want to advance, you have to do what most people don't ;-) and so you will get a hard life...
SDXL may be the last model that supports the "knowledge" of so many different artists/art styles and is much more capable of creating new styles by itself when combining those styles (just like most new artists get inspired too).
I have trained some loras to get a feel for what is possible. If I wanted the same result, I would have to train about 180+ Loras with the different artstyles of the ones I like and then experiment with the weighting - which will be a bit time consuming ;-). For me, v8 might be the best balanced for the styles I want to achieve. My lora folder has 200+GB of mostly unused loras - except for merging experiments.
What I can recommend is to use an AI like Claude 3.5 to generate some wildcards for dynamic promting, a good LLM can be quite innovative in mixing artist styles for new approaches (here, too, the right promting is essential). By using the promting capabilities of automatic1111/forge (i.e. switching parts of the promt after a given amount of steps) the possibilities are nearly endless. If you have enough time, comfy ui is a great tool with more options and influence on the result - but has a very bad ui and one of the worst user expirences... in my opinion.
Overall, I think v9 is exactly my cup of tea, as I tend towards photorealism.
And don't mind me complaining, I'm never satisfied. =)
You work too fast Zavy! I had my head full of tests including the V8... and tss V9, go DL! I'll try to do it faster before V10 lol
Considering what the images show, there is a good uptick in quality.
So, I went up to test my latest direct tests on the V9 ^^
I've been using V3 all this time. I was shocked to see it's on V9. Nine!
I love your models, they're by far my most used. The prompting is reliable and coherent, anatomy is pretty good most of the time, quality for realism and semi realism is outstanding! I often use it for inpainting and it works splendidly as well. Please keep up the great work, your work is extremely appreciated. I will be playing with this later tonight. V8 was epic!
Thank you for the nice comment!
Great!!!
Is there an inpainting version?
you can do it yourself
https://civitai.com/articles/833/create-an-inpaint-model-from-a-checkpoint-model
or use fooocus it generates automatic ;)
@sevenof9247 doesn't work for xl
@neuronsactivated for shure ! i use it like SD1.5 ... same options work in a1111 or webforg ;)
"DPM++ 3M SDE Exponential works "damn well, very damn well..."
You are the master of checkpoints! By far my number 1 go-to checkpoint since I started , at around v.5... Every time I run through my checkpoints with a prompt you always win. Thanks!!
And because of this huge quality gap from other checkpoints, it would be a dream if you would augment with some equally fantastic NSFW. Then I will only need one checkpoint tbh! <3
Thanks, such a nice compliment is very much appreciated!
With regards to NSFW, it's generally advised to get one of the many NSFW LoRAs and add it at a low weight, it should nudge the somewhat existing knowledge of NSFW in this checkpoint towards what you want. I don't personally want to overly focus on it beyond giving it the base capabilities where LoRAs hook into nicely.
I double that. You always win. Maybe RealismEngine gets somewhat close, but ZavyChromaXL is the best choice for photorealism and skin tone. Thank you and keep it on.
@Zavy I hear what you are saying.. it is just that most NSFW loras seems to be like more hardcore than aesthetic. Of course much of this genre is supposed to be, but I was just thinking with your exceptional SFW model it could be fantastic with your take on this, instead of just importing the sleazy stuff :D
What existing lora could be a good addition to your model at this point? of course this is not a professional question, just like artistic man to artistic man ;D
So actually I mostly think that your "nudeness blackbox" is a pity. For instance creating a topless woman, when a rare seed allows the toplessness, everything except the naked chest is created/modelled awesome, but the chest looks like it is importet from somewhere else :D
Interestingly, it is getting silent here, after the appearance of Flux.
I still get pretty much the same usage numbers on the generator. The chatter often comes for a few weeks when a new version is released. Which to be fair with Flux around may not be too useful, and feeling that this SDXL finetune is close to the limit of where I want it to be... it may not have many new versions coming.
@Zavy going to flux versions from now on?
@andux Primarily yes. I won't stop work on SDXL fully though. There may be an update or two in it yet. May also do some new Flux loras for SDXL as well. We'll see.
This model is amazing, the best SDXL model. Would be nice to have it ported to Flux.
Thank you! I will be working on a Flux version, but it's pretty expensive compared to SDXL, and there's not a great avenue of getting those costs covered beyond asking for donations with Flux's license. It's a bit of a downer.
another good model that is a hidden gem