Type | |
Stats | 107,226 |
Reviews | (4,320) |
Published | Jan 6, 2024 |
Base Model | |
Training | Steps: 1,620,000 |
Hash | AutoV2 AEB7E9E689 |
UPDATE:
Added Lightning Version for Version XI
For business inquires, commercial licensing, custom models, and consultation contact me under [email protected]
Join Juggernaut now on X/Twitter
Prompting Guide by Adam for XI & XII
Detailed Prompting Guide by Adam (Highly Recommended)
A big thanks for Version XI goes to RunDiffusion and Adam, who diligently helped me test :) (Leave some love for them ;) )
Hi everyone, long time no see ;)
As you may have already noticed, Juggernaut XI has been available on several platforms (NightCafe, Octo API) for the past few weeks. I could tell you a lot about it, but let's keep it simple. I’m now working on this full-time, which means I need to make money from it—not just to pay for the training (which I’m still funding myself) but also to cover my rent and put some food on the table.
This new release schedule gives me the opportunity to earn some money for myself and the team. The "downside" is that you had to wait a little longer for an open release. Personally, I think that's a small price to pay :)
Actually, XI was supposed to be released a bit earlier, but like many others, I fell down a Flux rabbit hole in the past few weeks
Now, a bit about Juggernaut XI: After our prototype with X responded well to the new captions, it was clear that we wanted to re-caption the entire Juggernaut dataset (15k). We’ve done that, and it's significantly more stable than version X. I still recommend a tag/keyword prompting style, but "normal" prompts should also work well :) I also tweaked the captions a bit before training.
Other than that, Juggernaut XI is still an SDXL model. I suspect expectations have risen quite a bit after the release of Flux. SDXL still suffers from some "issues" that are hard to fix (hands, faces in full-body view, text, etc.). But rest assured, we've tested it extensively over the past few weeks and, of course, compared it with older versions (X & 9). It’s even easier to prompt than version X and now has the visual quality of version 9 again.
Since the Juggernaut dataset contains only a small amount of nudity, male genitalia are again harder to prompt. For that, I recommend using either version X, version 4, or one of the many Penis LoRAs on CivitAI.
My personal recommendation is currently a Flux>Jug XI pipeline for the best possible output :) This showcase is RAW and not cherry-picked. A classic "What you see is what you get" ;)
Looking to the future: Juggernaut XII will soon be available on Octo and RunDiffusion (I will, of course, keep you informed), and some time after that, it will also be available here on Civit. I think my journey with SDXL will then come to an end, and we will focus on the next steps. In the past few weeks, we've had many tests, discussions, trainings, etc. Our main focus is currently clearly on Flux. When exactly we’ll release something here is still up in the air, but expect it within the next four weeks.
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. 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
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)