Updated the Prompting Guide
For business inquires, commercial licensing, custom models, and consultation contact me under [email protected]
Join Juggernaut now on X/Twitter
Juggernaut Ragnarok on RunDiffusion
Juggernaut XI & XII on RunDiffusion
Prompting Guide for Juggernaut Ragnarok by Adam
Prompting Guide by Adam for XI & XII
A big thanks goes to RunDiffusion and Adam, who diligently helped me make it work :) (Leave some love for them ;) )
Hey everyone,
It’s been 8 months since the last version was released here on CivitAI.
Of course, I haven’t been idle during that time . I completed several projects to ensure I’d have the financial means to keep exploring new architectures and possibly do full finetunes on them in the future.
Juggernaut Flux (and its many sub-variants) was a ton of work, but ultimately, I’ve wrapped that chapter up. The training process gave me way too many headaches. To keep my sanity, I spent my spare time working on Juggernaut SDXL with the hope of maybe releasing one final version for you all.
And that day has finally come. :)
Juggernaut Ragnarok has improved in many areas: photorealism, digital painting, poses, hands, feet, and much more.
That said, it’s still an SDXL model, and I don’t recommend comparing it to models like Flux, Reve, or Sora. For example, it still has limitations when it comes to text rendering or faces at a distance.
I recommend using it as part of a pipeline for your projects. Example setup:
FluxDev / Pixelwave / Jug Flux Pro → Juggernaut Ragnarok
A quick personal note about Juggernaut:
Honestly, I don’t know what comes next.
After the release of Sora and similar tools, the open-source image generation space feels a bit dull in comparison.
Nothing has really excited me enough to dive back into training (yes, I’m talking about HiDream too).
So I’m seeing Juggernaut Ragnarok as a kind of farewell, especially since it’s unclear where things are headed with CivitAI in general.
(You can download all Juggernaut versions from HuggingFace, by the way.)
Last but not least:
Have fun with the model, share your creations, and good luck with your projects!
And in case you’re wondering: Yes, you can do anything you want with Juggernaut : merge it, train it, sell the image outputs, etc.
Just a simple shoutout is all I ask. :)
And now, here are the recommended settings:
Recommended Settings(VAE is baked in):
Res: 832*1216 (For Portrait, but any SDXL Res will work fine)
Sampler: DPM++ 2M SDE
Steps: 30-40
CFG: 3-6 (less is a bit more realistic)
Negative: Start with no negative, and add afterwards the Stuff you don´t wanna see in that image.
VAE is already Baked In
HiRes: 4xNMKD-Siax_200k with 15 Steps and 0.3 Denoise + 1.5 Upscale
And now, have fun trying it out. As always, I'm eagerly waiting for your pictures in the Gallery :)
If you liked the model, please leave a Like. In the end, that's what helps me the most as a creator on CivitAI. :)
Last but not least, I'd like to thank a few people without whom Juggernaut XL probably wouldn't have come to fruition:
Dreamlook.AI (Trained 3 Side Sets)
Description
The whole Juggernaut 9 Dataset recaptioned and trained for XI (15k Images)
Res: 832*1216 (For Portrait, but any SDXL Res will work fine)
Sampler: DPM++ 2M SDE
Steps: 30-40
CFG: 3-6 (less is a bit more realistic)
Negative: Start with no negative, and add afterwards the Stuff you don´t wanna see in that image. I don´t recommend using my Negative Prompt, i simply use it because i am lazy :D
VAE is already Baked In
HiRes: 4xNMKD-Siax_200k with 15 Steps and 0.3 Denoise + 1.5 Upscale
FAQ
Comments (74)
Working on it :)
@KandooAI fp8 and nf4 bnb if possible please
@NoobFromEgypt nf4 ..nope .. better Q4
Gonna try Version XI, can't wait to see what you did there! Oh and I am excited to see what you're gonna do with FLux, as the Dev version is non-commercial and you need to earn I guess it's gonna be based off of Schnell? I hope so...
@SirMeon I can´t talk about Details yet. But we already had some Good Talks with Black Forest. Right now i am thinking we gonna train the Dev Variant and get a License for it.
Schnell Version is good, but much worse compare to the Dev Variant. And of course its highly distilled, that makes training way harder
@KandooAI Если всё получится и juggernaut объединится с flux получится всемогущее создание. Я буду этому очень рад.
@KandooAI Would that mean if I wanna sell artworks generated with your JuggerFlux Model, I would have to buy a license as well? 🫤 I understand the quality and training aspect though, of course!
@SirMeon No, as an artist selling your own artworks created with a JuggerFlux model, you wouldn't need to buy a license. The licensing is more aimed at businesses, like Generation Services, StockPhoto Sites etc. that use the Model in their Workflow
Added a new Prompting Guide for XI and the upcoming XII. The Guide is made by Adam, so a big thank you to him :)
https://learn.rundiffusion.com/prompt-guide-for-juggernaut-xi-and-xii/
Please add a hyper version for XI when you have time.
Don´t worry i will be uploading it somewhere next week :)
@KandooAI hello, any update on a hyper version of XI? it's been 13 days since your message ;-)
@mikrofu242 just use the lora for the moment, same thing.
@cptdan ok!
@mikrofu242 I am currently sick and lying in bed with a flu -.- So it got a bit delayed. But i am optimistic that i can upload it by end of the week :)
@KandooAI no worries mate, we hope you recover quickly and feel better soon. Take care and get plenty of rest and fluids!
@KandooAI ok get well ! take it easy !
I'm having much better results now with XI than X. Thank you for releasing this new version, I'm loving it so far.
Nice ;)
Can't wait a Flux dev form you!
Thanks for another great update! BTW, In the description of the new Juggernaut XL model it says that the best settings are "DPM++ 2M Karras", but all the preview images are made with "DPM++ 2M SDE". So which one should we use?
Thanks for the Hint...I forgot to change that in the Description....I recommend DPM++ 2M SDE, but of course you can use others too
thank you, I am getting much better results with DPMPP_2M rather than DPMPP 2M SDE. SDE seems to distort a bit but has better prompt adherence. Great update tho
I don't personally use DPM++ 2M at all, but when I do I use SDE Karras, so both at the same time via sd.next. When using SDXL in general I've found the best details and textures come from DPM SDE Karras or DPM++ 3M SDE Karras. Anecdotal.
Generally, SDE gives better details and more creativity/hallucination. Non-SDE has better consistency and adherence. It's a trade-off, and mostly just personal preference.
what's the difference for Jugg_XI_by_RunDiffusion and Jugg_XI_by_RunDiffusion_Hyper?
Versions of a checkpoint that have "hyper" in their name can be used with less steps for great results. You are trading a bit of quality for a lot of time. Sometimes you want that image fast - or just don't have the VRAM for the big models to work in reasonable times.
where is Jugg_XI_by_RunDiffusion_Hyper?
Tried inpainting with it in Fooocus and it gives horendous results for some reason?
I think I know why, this model doesn't play well with the Fooocus Enhance Style.
@joelraechamp85264 Thanks for the tip! I was getting very strange cartoon-like images for some prompts. Substituting Fooocus Semi Realistic for Fooocus Enhance fixes the problem. With Pony models you have to do the same thing.
me too. so i deleted it .
Gives definitely better results than the previous version. Great improvement! Nice job! Congrats.
As a non-native English speaker, reading Adam's prompt guide made me feel like I'm relearning English. So many intricate wordings! (Nice guide btw, I love it, learned both SD and English with it)
You've discovered purple prose! Don't worry, no one actually speaks like that.
How can I implement "HiRes: 4xNMKD-Siax_200k with 15 Steps and 0.3 Denoise + 1.5 Upscale" Into ComfyUI? Can Anyone give me a sample workflow?
Use Ultimate SD
Jugg_XI_by_RunDiffusion_Hyper , where is it ?
Please write a universal workflow for: Juggernaut xl rundiffusion. I have ComfyUI.
What sampler? What CFG value? How many steps? What universal words do you use for positive prompt? What words do you use for negative prompt? I just render all images on my processor, not on the video card, since my video card is from AMD. Rendering ONE image takes me at least 20 minutes and it is 512x512, and since this is an XL model I will have to use a resolution of 1024x1024 and this takes me at least an HOUR.
Workflows are all over and Juggernaut doesn't require anything special for it to work. The recommended cfg, sampler and steps are already on this page. Juggernaut is a great model but you also need to do a bit of work. All of this is for free remember.
And to add to what GodAIMighty said; there's literally a prompting guide linked in the description of the model. Try reading all the available information before getting pissy.
Hey man did you check out my Prompt Guide I have all the settings spelled out there. recommended settings are definitely not 512x512... you should hit up SD1.5 models for that resolution or consider cloud services like Nightcafe, RunDiffusion etc.
all the answers to the sampler, steps etc are in the description. if it takes an hour to render, then probably any other activity in the world would be more worth the time.
impressive! Nice update
could you explain your Flux/Juggernaut workflow?
Is it generation in Flux and IMG/IMG in Juggernaut with low denoising? Thx
Yeah pretty much how you describe it....When i use a Flux/Juggernaut Workflow i generate the Image first with Flux and then use the Image for the Img/Img Process with Juggernaut. I prefer using 0.30 Denoise
It's really good for general use or mixing with loras.
Simply amazing! Demonstrates incredible power and effectiveness. Great job!
Enjoying XI quite a lot, it follows prompts really well.
Amazing
can you tell me what trainer used for this model?
Kohya_ss
ok your model is looking good with 15k dataset but I had trained with 84k large dataset but did not get good output what method do you use for this model
how can i create less square and angular faces? every time i try to make an image on fooocus with JaggernoutXI i only get angular faces T-T
on fooocus I can't even use the XI version, most times I get a monocolored silhuette and that's it...
@naughtyskynet i just saw the reply, sorry. What do you use then (Automatic1111 or others)?
wonderful, thank you
Nice!
Unfortunately, no one answered my question about universal settings normally, but I was still able to find them (without Lora) (I have comfyui):
Clip-Skip: -2.
Resolution: 1024x1024.
Steps: 30.
CFG: 6.5.
Sampler: dpmpp_2m, karras.
Positive: Skin Textures, High Resolution, Cinematic, detailed texture, hyperdetailed, high dynamic range, photography, sharp focus,
Negative: (deformed, distorted, disfigured:1.3), poorly drawn, bad anatomy, deformed iris, (bad hands:0.7), wrong anatomy, extra limb, missing limb, floating limbs, (mutated hands and fingers:1.4), disconnected limbs, mutation, mutated, ugly, disgusting, blurry, amputation
Great model
Question for the model creator. So what full settings do you usually use? The settings that are in the examples are not universal at all. Yes, close-ups are fine with them, but if I want to move the camera a little further or just see the character's hands, everything breaks. Just please don't send links to any instructions and so on, where everything is very abstract (cfg 4-6 is a very strong spread), and changing just one parameter in a neural network can change a lot. Yes, of course, I picked the settings, but I really feel that there are better ones that no one has written. I don't want to fix the model with a bunch of Lora and embeddings.
True 👍🏻
Great model. Thank you!
Very impressed with your work.
I'm comparing 9 vs current...
would be nice if the release notes for 9 were somehow viewable?
I dont see how to view them :(
Used it with Stability matrix, its great
I can't get it to create full length body images
mention feet/shoes or the type of floor they're standing on
Narz's advice is best. Also the aspect ratio of your image can impact this, set it to portrait but do not mention words like portrait because it causes close-ups
can u tell me what CLIP SKIP value i have to set? and is better to use DPM++ 2m sde or DPM++ 2M karras or eleuler ancestral?
Prompt Guide (rundiffusion.com)
read this
Amazing work!
Is it me or does it feel like this model has become poorly aged in ReForge latest build, getting XI in hopes of getting a reduction in bad hands failures.
Ever since Forge June update broke everything ReForge has been unpredictable at best
Details
Files
juggernautXL_juggXIByRundiffusion.safetensors
Mirrors
juggernautXL_juggXIByRundiffusion.safetensors
juggernautXL_juggXIByRundiffusion.safetensors
juggernautXL_juggXIByRundiffusion.safetensors
juggernautXL_juggXIByRundiffusion.safetensors
juggernautXL_juggXIByRundiffusion.safetensors
juggernautXL_juggXIByRundiffusion.safetensors
juggernautXL_juggXIByRundiffusion.safetensors
juggernautXL_juggXIByRundiffusion.safetensors
juggernautXL_juggXIByRundiffusion.safetensors
juggernautXL_juggXIByRundiffusion.safetensors
juggernautXL_juggXIByRundiffusion.safetensors
juggernautXL_juggXIByRundiffusion.safetensors
juggernautXL_juggXIByRundiffusion.safetensors
Juggernaut-XI-byRunDiffusion.safetensors
juggtXL_juggXdiffusion.safetensors
ae-pixel-xl-v2.safetensors
sdxl-juggernautXL_juggXIByRundiffusion.safetensors
Juggernaut XI by RunDiffusion.safetensors
SDXL_Juggernaut-XI-v11.safetensors
juggernautXL_juggXIByRundiffusion.safetensors
juggernautXL_juggXIByRundiffusion.safetensors
Juggernaut-XI-byRunDiffusion.safetensors
juggernautXL_juggXIByRundiffusion.safetensors
juggernautXL_juggXIByRundiffusion.safetensors
juggernautXL_juggXIByRundiffusion.safetensors
Juggernaut-XI-byRunDiffusion.safetensors
juggernautXL_juggXIByRundiffusion.safetensors
Juggernaut-XI-byRunDiffusion.safetensors
juggernautXL_juggXIByRundiffusion.safetensors
juggernautXL_juggXIByRundiffusion.safetensors
juggernautXL_XI_Rundiffusion.safetensors
Juggernaut-XI-byRunDiffusion.safetensors
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.