Pruned Version 7 is now available
Use this VAE : vae-ft-mse-840000-ema-pruned
Info:
Hello dear friends,
I apologize for the silence regarding my models lately. I’ve been dealing with serious health issues and spent a long time in rehab, where I had to relearn how to walk properly. As if that wasn’t challenging enough, when I finally got home, I noticed something was wrong with my PC. Shortly after turning it on, there was a small bang—apparently, my water cooling system malfunctioned due to the long downtime, and it leaked onto my hardware.
Now, I’m faced with the task of completely rebuilding both my databases and my PC. Unfortunately, due to the costs of rehab, I’m not in the best financial situation, so I’ll have to restore everything little by little.
Still, I hope Version 7 manages to impress you and brings you joy while generating images and more.
Version 7
Detailed face
Better generation with hands
Detailed Body
Better Shadow
trained size 768x768, 1024x1024
When creating the model this time more attention was paid to the aesthetics as well as multiple people, and that it looks even more real
Description
In V1 the limbs are sometimes created too long, I'm working on it now so that is fixed
FAQ
Comments (29)
Excuse me but what does it mean "different species"?
Oh yeah - did you not realise that women come in different species: ginger, cute, thick, and so on... 😂
I apologize for the confusion. I have considered the various ways in which the body can move and deform. This includes a wide range of positions in which people can be photographed. It also includes the different angles of the photos, close-ups and long shots, the lighting conditions, and the nationalities of the people in the photos.
@Haakii Ah ok thanks, mate. Understood.
Including men?
@aik are men a species?
@squidman 😂😂😂😂
@aik By merging other models, male images are also possible, 80% were female when created and thus 20% male images were present during refinement
@SerenityX :2.4 for the people in the back
Looks interesting... Probably worth listing what was in the merge 🤔
Here is a list of the steps I took to create the model
First, I created an extensive base by reading in over 5000 photos of people of various nationalities, both clothed and nude.
Then the first model, AmiReal, was used to generate images.
After determining that the shadow cast was not correct when generating the people, I used over 500 images to refine the shadow cast on bodies and objects.
Then I used the next model, CyberRealistic, and combined the two models.
Unfortunately, this combination resulted in either strange faces or the same faces over and over again.
So I trained the model again and took over 3000 close-up shots of faces until the result was acceptable.
For my last combination, I decided to use the realisticVision and realhotMix models.
Then I tried again to refine with 500 images only hands and feet.
@Haakii Yeah, that figures! AmIReal, CyberRealistic and RealisticVision are three I use regularly for realistic images... ❤️ Thanks for the info too!
@SerenityX am i perfection v2 is also great at them too in my humble opinion
Please give more information about what this model is specialized in and what it's not good at
Here is a list of the steps I took to create the model
First, I created an extensive base by reading in over 5000 photos of people of various nationalities, both clothed and nude.
Then the first model, AmiReal, was used to generate images.
After determining that the shadow cast was not correct when generating the people, I used over 500 images to refine the shadow cast on bodies and objects.
Then I used the next model, CyberRealistic, and combined the two models.
Unfortunately, this combination resulted in either strange faces or the same faces over and over again.
So I trained the model again and took over 3000 close-up shots of faces until the result was acceptable.
For my last combination, I decided to use the realisticVision and realhotMix models.
Then I tried again to refine with 500 images only hands and feet.
@ChangeMeNot you should post this in the description
what size is it trained in?
It might be ok for vanilla pics, but adding a girl and a man with his wang anywhere near her, causes issues without a lora. And let's be honest, won't she be lonely without a wang? :)
Been playing with it and so far it works great, with and without Loras. Will post some images once done.
Good model, very flexible and plays nice with multiple LORAs and TEs. Impressive.
Model is 5.28 GB. Model type identified as SD-v1. Model components are: UNET-v1-SD, CLIP-v1-SD, VAE-v1-SD.
Contains no junk data. Wastes 3.30 GB on precision.
Uses the SD-v2 VAE.
Model will be pruned to 1.99 GB.
size can be cut down to 2gb FYI
Even with the precision, all people look like Asians standing in a dark room.
Looks great, gonna wait for a prunned version
im impressed looking forward to more keep going!
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.