(2024/03/01) OhMen Origins V3 is online !
OhMen Origins is the heir of Anymen (one of my very first models), and is the trained mix used in OhMen Lifelike.
You can still check my previous mixes here.
>>> Please, make sure to check out the model description, especially the "best settings" and "to go prompts" sections <<<
OhMen Origins is specifically designed to enhance male characteristics by combining various base models.
This unique combination allows for the effortless creation of digital paintings, 2.5D, and 2D images featuring attractive males and a unique style, without relying solely on LoRAs.
However, if you prefer, you can still utilize LoRAs as this trained mix offers high versatility. Additionally, OhMen Origins offers a diverse range of sub-styles, providing even more options for artistic expression.
If you're interested in bringing OhMen Origins to life through merging, you can refer to the "Merge a OhMen model" section. This will guide you on how to combine and blend different elements to create your desired result.
╭───────────────────────╮
│My to go prompt
╰───────────────────────╯
Positive prompt : (1man), brown eyes, handsome,
Negative prompt : FastNegativeV2,deformityv6, bad-hands-5, (jpeg, deviantart, worst quality:1.2,low quality:1.2) ,╭───────────────────────╮
│Best Settings
╰───────────────────────╯
Clip 1 is highly recommended, while Clip 2 will likely generate pics with less details.
DPM++ 2M Karras, DPM++ 2M SDE Karras, or my personal favorite, DPM++ 3M SDE Exponential Sampler. Restart works well too
For sampling, try using 35-60 steps.
To get the best results, set your CFG scale at 6. It works great between 3.5 and 8.
Make sure to uncheck the Face restore option.
Recommended resolutions fall between 512x512 and 768x768.
You can go for 1024x1024, but sometimes it can generate some weird stuff.Keep in mind that upscaling can sometimes reduce the level of details instead of increasing it.
To counter that, I recommend using a denoising strength of 0.35 or 0.4.
Also, try using 15-20 Hires steps and either "4xNMKD-Superscale-SP_178000_G" or "RealESRGAN 4x+ Anime6B" as your upscaler.if you want your faces to look super clean and realistic, you should totally give the AfterDetailer extension a try. It's not mandatory since OhMen usually does a really great job with portrait, but using it can add that extra touch of realism to your composition. Same for penises, but you can check out this Penis model for the AfterDetailer extension ;
The Vae-ft-mse-840000-ema-pruned is included but you may need to set your VAE on "Automatic" ;
If you wanna gen some NSFW pictures , "flaccid penis, nude, saggy balls, ..." are my to go NSFW tags.
╭───────────────────────╮
│Known Issues
╰───────────────────────╯
V1
Feet aren't properly generated : SD doesn't generate feet properly most of the time. However, you can solve this issue by adding the "Neg Feet" embedding in your negative prompt or using a LORA such as the "Male Feet Pose" one.
V2
You don't like beefy or super hairy guys ? Add "(Muscular:1.5) or "(hairy:1.5)" in your negative prompt, I'm gonna try to fix that on the next update.
V3
╭───────────────────────╮
│Merging a OhMen model
╰───────────────────────╯
A OhMen model represents the culmination of countless hours dedicated to make it stable and easy to work with. So, I recommend the following:
Set OhMen at 0.5-0.6 as Checkpoint A (if you plan to use it as a base model) or 0.35-0.5 if OhMen is your Checkpoint B.
If you plan to merge LoRAs, merge them with higher values before adding OhMen to your mix.
As OhMen already includes HQ built-in penis, set a lower value if you are merging one.
In case you are block merging and only want a specific feature from the model (as checkpoint B) :
Strong overall style, textures affected: 0,0,0.5,0.5,1,1,1,0.5,1,0.5,0.5,0.5,0,0,1,1,0,0,0,0.5,0.5,0.5,0.5,1,1
Strong overall style and composition, genitals untouched (from a Rentry post): 1, 0.9, 0.7, 0.5, 0.3, 0.1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0.1, 0.3, 0.5, 0.7, 0.9, 1
Alternative overall style and composition: 0,0.9,0.7,0.5,0.3,0.1,1,1,0.35,1,1,1,0,0.45,0.25,0.25,0.5,0.35,0.5,0.25,0.3,0.5,0.7,0.9,1
strong overall style and mid-weak composition: 0,0,0.7,0.5,1,1,1,1,1,1,1,1,0,0,1,1,0,0,0,0.1,0.15,0.3,0.5,0.9,1
Mid overall style and composition, genitals slightly affected : 0,0.5,0.5,0.5,0.5,0.5,0.5,0.5,1,0.5,0.5,0.5,0,0,1,1,0.3,0.5,0.3,1,0.5,0.5,0.5,0.8,0.8
weak overall style and composition, genitals untouched: 0.2,0.5,0.5,0.5,0,0,0,0.3,0.45,1,1,1,0,0,0.45,0.4,0,0,0,0.25,0.35,0.4,0.35,0.7,0.7
Genitals, slightly affects facial features: 0,0,0,0,0,0,0,0,0,0,0,0,0,0.3,0,0,1,0.55,1,0.25,0,0,0,0,0
Strong genitals, skin textures and faces affected : 0,0,0,0,1,1,0,0,0.5,0,0,0,0,0.5,0.5,0.5,1,0.55,1,0.45,0,0,0,0,0
hand repair (based on a Rentry post)(may requires a top-rated or well trained model): 1,0.6,0.40,0.20,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
╭───────────────────────╮
│Credits
╰───────────────────────╯
A HUGE thanks to these testers :
Karya, Uranian, AI4Fun, Valen Pyke, zalkalwe404, and Czach (V1)
DigitalGhosts https://civarchive.com/user/digitalghosts
Resonance https://civarchive.com/user/resonance_ai
Boomerang https://civarchive.com/user/b00merang
Revelancy https://civarchive.com/user/Relevancy
Markury : https://civarchive.com/user/markury
Checkpoints:
V3 has been reworked upon the updated base models from Origins V2
- OhMen Origins v2
- DreamShaper : https://civarchive.com/models/4384/dreamshaper
- Indigo Furry Mix : https://civarchive.com/models/34469/indigo-furry-mix
- AnythingQingMix : https://civarchive.com/models/85676
- HardCandy : https://civarchive.com/models/199409/hardcandy
- EroGen : ....
The whole mix has been trained with male pics from Midjourney Niji
Loras:
(V3)
OhMen fusion : Personal Lora , from a combination of some beta and alpha OhMen models
(V2)
- DPO Lora : https://civarchive.com/models/242825?modelVersionId=273993
- Furthermore : https://civarchive.com/models/157105/furthermore-detail-enhancer
- Stock Male MJ : Personal Lora , trained on male stock-like photos sourced from my Midjourney RAW prompts
- Anymen RD : Personal Lora , from a combination of Anymen and Midjourney pics
(V1)
- tangbohu detail maker : https://civarchive.com/models/62687/tangbohu-detail-maker-lora-lora
- Anymen style : Personal Lora
Please, note that with OhMen Origins, you have the ability to sell the images generated using this model. It provides you the opportunity to monetize your AI artworks :)
Airfuck , KingApollo and Muserato ❤️💙
Description
Base checkpoints updated and block merged with the trained OhMen Origins mix