Check my models on Mage and get unlimited generations: ConsistentFactor / Consistent Factor Euclid / Consistent Factor Borealis / Consistent Factor Lumina / Consistent Factor Allura / Consistent Factor Keter / CFXL2 / LobeliaXL 2 / AuroraXL / AuroraXL+ / ArtemisiaXL / FuchsiaXL / PetuniaXL / DhaliaXL / DhaliaXL 2 / CFXL 3 / CFXL 4 / LobeliaXL 3 / AzaleaXL / LotusXL / CFXL 5 / FuchsiaXL 2 / DaisyXL / NolanaXL / AjugaXL
What is Euclid?
Euclid is a merged model trying to get as close to realistic as I can. Currently, there are a number of models in the merge, with version 6 being the most recent official release. Version 6 increases the realism to a new level, not only that, there is an Ultra version which adds more contrast and details whilst fixing the dreaded weird eye problem.
One thing I'm happy to celebrate about my model as that it is a lot more racially friendly and will more often than not, produce people who are not Caucasian, in fact around 60%-70% of the time it will produce people of colour in my testing at least. If you find it's not producing Caucasian people, just add either white or Caucasian to the prompt. It's also very good at aging people so adding an age can make a big difference.
The only thing V5 doesn't do well most of the time are eyes, if you don't get decent eyes try adding perfect eyes or round eyes to the prompt and increase the weight till you are happy. V6.1 and V6.1 Ultra have fixed this problem.
What is V6.1?
V6.1 is the final step for Euclid. Yes, that's right, this is the final version. Months of hard work, learning, merging, all to get to this point. There will be side versions like Ultra and Cinematic, but this is it as far as the base version goes. I've pushed SD 1.5 as far as it will go. I hope you all enjoy.
What is V6.1 Ultra?
%2C%20Steel%20gray%20hair%20loose%20braided%20hair%2C%20comics%20printed%20shirt%2C%20oversized%20jogger%20pan.png)
V6.1 Ultra takes the foundation of V6.1 and enhances it. You can expect more realistic colours, better contrast, and better texture. As well as fixing the dreaded weird eye problem that has plagued my models in the past. If you are looking for the most realistic version then I would highly recommend this.
What is V6.1 Cinematic?
%2C%20sitting%20on%20the%20floo.png)
V6.1 Cinematic is exactly what it sounds like. It's a more cinematic and sometimes dramatic version of V6.1 Ultra. So expect enhanced lighting and shadows as well as a more dynamic feeling image. It does sacrifice some of the realism from V6 Ultra to achieve this but I feel the trade-off it worth it.
Recommend for V6.1: Huen or DPM++ 2M SDE Karras, Steps: 20-45, Hires fix 0.15 - 0.4, CFG 6 - 8.
I would also recommend using epi_noiseoffset for V4 - https://civarchive.com/models/13941?modelVersionId=16576
I would recommend using Theovercomer8's Contrast Fix for V5 - https://civarchive.com/models/8765?modelVersionId=10638
I would also recommend using difConsistency RAW VAE
Description
V6 Ultra takes the foundation of V6.1 and enhances it. You can expect more realistic colours, better contrast, and better texture. As well as fixing the dreaded weird eye problem that has plagued my models in the past. If you are looking for the most realistic version then I would highly recommend this.
FAQ
Comments (10)
WOW! That is the only thing I have to say about this model. Truly incredible work. It is so coherent, nothing even seems to janky. The only stitch I have with this model is that the colors seem a little dull sometimes. But incredible
For Ultra, what is the difference between the 2GB and 5GB downloads? same name, different sizes
Recommended hires 0.15, but what setting is that referring to? Denoising? And are you recommending it even if not actually upscaling?
I would also like to make a realistic model, and I'm currently trying through Dreambooth spending a lot of time on it, but I can't figure out the settings, the results are disgusting. You can tell me what settings you use, what resolution of images and according to what scheme the training takes place. On the Internet, I could not find lessons on teaching such large-scale models with a package of images of several thousand. I trained 1900 portrait images of 512x512 at 380000 steps tonight (Learning Rate 0.000001), as a result, some kind of crooked nonsense turned out. Save me, brother.
Is the nearly 6GB file SDXL?
Do you want the next version of Euclid? Or maybe the soon-to-be-released CFXL (coming this weekend) and AuroraXL (coming in the new year)? Then head on over to the only place you'll be able to get my models https://consistentfactor.vsble.me/
Does anyone know where we can get keter?
Hey, when generating images with your CFXL model what do I need to do to make images sharp?
They are mostly blurry or like out of focus, not showing much details?
It's the one I just keep coming back to.
A nice model. Good variability and responsiveness, very good photorealism, normal anatomy.
Good job!
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.