If you'd like to buy me a beer or coffee! 🍺☕
https://ko-fi.com/scratchproof
Virile Reality V3 BETA 2
Rebalanced models and loras
I would advise to use Clip Skip 1 for more realistic results (Thank you b00merang)
Virile Reality V3 BETA
Because who needs XL?
New mix
Better flaccid dicks
More realistic
Faces should be better and sharper
What's included:
Checkpoints:
Virile Reality V2
Epic Realism - Natural Sin
Unstable Homoerotic Diffusion
Juggernaut Aftermath
Personal Checkpoint
Homoerotic
Virile Fantasy
Loras:
T-Girl
Better Cocks V2
Personal Lora (Softeepee)
Penis V2
Humans
Add-Detail
Virile Reality V2
Rebuilt from the ground up!
Now with flaccid penises (prompt for "flaccid penis")
New merges:
A-Zovya Photoreal
Unstable Homoerotic Diffusion (UHD)
SofteePee Lora (Own creation, not available as a standalone yet)
BetterCocks Flaccid
Instant photo 拍立得/Polaroid LoRA
My settings (A1111):
Sampling Method: DPM++ SDE Karras
Sampling Steps: 28
Hires. fix On
Upscaler: R-ESRGAN 4x+
Hires Steps: 14
Denoising Strength: 0.5
Upscale: 1.75
Width: 512
Height: 768
CFG Scale: 8
I also highly recommend using After Detailer for faces, using the "mediapipe_face_full" model with the "Restore faces after Adetailer" option enabled.
Hope you enjoy!
This model is my second merge and has been fine-tuned in order to produce convincing and realistic male-focused gens without the need for Loras, even with simple prompts.
This model is built upon my previous merge, Virile Fusion, however it is not a replacement and only aims at offering a more realistic style, which could not be achieved by VF.
In the end, whichever model you decide to use is a matter of preference.
Models included in the merge are:
Virile Fusion
Realistic Vision V2.0
ICBINP - "I Can't Believe It's Not Photography"
Juggernaut
Loras included in the merge are:
Penis LoRA
Tgirls
While the model has been crafted to be male-centric, it also allows for the generation of other types of characters (including females) as well as landscapes, or objects.
Recommended settings for good gens (Automatic1111):
Sampling Method: DPM++ SDE Karras
Sampling Steps: 22
Hires. fix On
Upscaler: R-ESRGAN 4x+
Hires Steps: 10
Denoising Strength: 0.5
Upscale: 1.7
Width: 512
Height: 768
CFG Scale: 6
I also highly recommend using After Detailer for faces, using the "mediapipe_face_mesh" model.
I did not bake any VAE into the model and would advise to use the following one: vae-ft-mse-840000-ema-pruned
I would also like to mention and thank the Unstable Diffusion Discord Server's #men-only community who helped me learn a lot. Some of the gens done with the model include their prompts.
Hide