Support the Freedom at https://ko-fi.com/lorian0x7
This model is also available on my kofi page without the need to pay with Yellow Buzz!
š“ MOSS :SHARP AND UNCENSORED
TIRED OF BLURRY / SOFT OUTPUTS?
MOSS checkpoints are notoriously soft and lack fine detail straight out of the box.
To solve this, I created a dedicated Deblur Workflow that restores crisp skin textures, sharp facial features, and clean edges without altering the generation style.š Get the Sharpening Workflow on Ko-fi: https://ko-fi.com/lorian0x7
The workflow it's set to work with an uncensored text encoder to improve prompts without the usual limitation of the default one.
ā¤ļø Support Independent Development
I am passionate about creating open tools, versatile checkpoints, and unrestricted models for the community. If you value this work and want to support independent development, please consider joining my supporters on Ko-fi:
š https://ko-fi.com/lorian0x7
š„ Available on Ko-fi:
Alpaca0x7: New Pony-based checkpoint with high-end anatomy handling.
MOSS Sharpening Workflow: The dedicated setup to eliminate the native blur of MOSS models.
Clothing removal Flux Klein Lora: Made to keep the body proportions and shapes as close to the original image as possible.
Early access to upcoming releases and experimental workflows.
š¦ Checkpoints Overview
Resilience (NEW - MOSS): Realistic freedom is back with this amazing MOSS checkpoint, extremely flexible and optimised for NSFW generations. No Turbo LoRA needed.
Omega: The flagship release. It delivers exceptional realism and artistic depth that rivals FLUX, maintaining high visual fidelity across complex scenes.
Ophelia: Focused on natural beauty and refined female anatomy while maintaining strong flexibility for everyday generations.
Aria: Soft aesthetic, exquisite lighting, and optimized for detailed portraits.
Wonderland: A solid, versatile everyday checkpoint with balanced anatomy, expressive faces, and rich lighting.
Alpha: The original SDXL base checkpoint with a cinematic, unrestricted touch.
āļø Generation Settings
š¹ Resilience (MOSS)
Steps: 8
CFG: 1
Sampler: Euler
Scheduler: bong_tangent or Beta
Note: Turbo LoRA not needed.
š¹ SDXL Checkpoints (Omega, Ophelia, Aria, Wonderland, Alpha)
Samplers: DPM++ 2M / DPM++ SDE 3M
CFG Scale:
2.0 ā 5.0 for short prompts
4.0 ā 9.0 for detailed prompts
Disclaimer: The creator holds no liability for any images produced with these models, including their legality, ethical implications, or potential misuse. By using these files, you agree that all responsibility for generated content rests solely with you. Use responsibly and within legal boundaries.
Description
New version, better, woman anathomy, more NSFW, better realism.
FAQ
Comments (14)
OK.... i am impressed. Dynamic Prompts, a slf-generated promptlist with 1000 more generic prompts and an own Lora, and this is so so good. Styles, Colors, Vibrance.
Well Done !
Dunno if it is just me, but it appears this checkpoint broke my install of Forge. Ever since installing and loading it the first time, I get:
TypeError: 'NoneType' object is not iterable
every time I go to create an image. This happens even when I switch to other checkpoints and after a complete system reboot.
I'm sorry to hear that, I don't have any issue using it with forge, maybe try with an update.
@LorianĀ From the sheer number of positive reviews and your own statement that it works fine on Forge, let's assume it is just me/my particular setup. I reinstalled Forge and will try again as the praise you are getting for this makes it seem like it is worth the effort. I'll reply if I don't somehow manage to screw the whole thing again. LOL.
@Inland389Ā Let me know, Otherwise you could try with A1111
I Use forge, this isn't broken - it's a forge bug. Change to one of the more standard resolution sizes. like 1024 square, 812x1236, etc. Also, make sure to clear out any overrides, sometimes those are the cause.
99% of the time if I do these two things it will render. (Forge is amazing, so I wouldn't want you to not use it.)
Not sure you can call this model "SFW and NSFW" given that it really really wants to generate nudes even though the prompt is totally safe for work, and even with negative prompts like "nsfw, nude, naked".
THere are different version of this model available. Ophelia is the most explicit one, you may want to try Wonderland.
Also, It's all about prompt engineering and It depends what you have in the positive prompt too. For example if you have "big boobs" in the positive prompt it can be a more powerful token then nsfw in the negative prompt try avoiding describing body parts in the positive prompt.
@LorianĀ Thanks for the clarification! And sure, I know that the pos prompt of course has an effect but I didn't have anything like that in the prompt that might have pushed the model towards nsfw.
I'm pretty sure the only way to get SFW results is to prompt empty landscapes. Any human, animal, or remotely animate object will have a giant minge.
This is a great model to produce beautiful images. So I did what I always do and tried to turn it on it's head. I preset to you The ButterFace Test
https://civitai.com/posts/2769165
Pretty good results!
very versatile, good results, well trained model
i just busted this out again after months of it being on the shelf, and immediately was reminded why I hung onto it. vastly underrated model