HubbaHubba was created for photorealistic women with and without clothing. Women are beautiful.
Auto1111webui is recommended to use this model.
This is the one I've used as it installs everything easily and even allows you to do command line arguments in the start program UI.
https://github.com/EmpireMediaScience/A1111-Web-UI-Installer
Should work with the normal v2 yaml.
It does some men shots well too.
Put penis in negatives for better female anatomy.
Merged with multiple 2.1 checkpoints and my own trained model and lora to create the most realistic women I could possibly manage. Most props go to the waifu 2.1 model which made training this much easier.
https://civarchive.com/models/17332/realism-engine-meet-waifu-diffusion
https://civarchive.com/models/21275/the-objectifier-34e
https://civarchive.com/models/22605/kitchensink2nsfw
https://civarchive.com/models/17307/waifu-diffusion-15-beta-2-sd21
https://civarchive.com/models/15975/artius-v21-nsfw
Made this as free to use and I'm not looking for any profits or donations.
Enjoy!
Description
1.0 version
FAQ
Comments (5)
config file .... hubbahubba_v10.yaml error.....
v2-inference-v.... good...
I get this error even with the setting turned on, it just doesnt work:
NansException: A tensor with all NaNs was produced in Unet. This could be either because there's not enough precision to represent the picture, or because your video card does not support half type. Try setting the "Upcast cross attention layer to float32" option in Settings > Stable Diffusion or using the --no-half commandline argument to fix this. Use --disable-nan-check commandline argument to disable this check.
if your using auto1111 go to the stable-diffusion-webui folder and open launch.py with notepad on line 13 it will say ('COMMANDLINE_ARGS', "")
inside the quotation marks try ('COMMANDLINE_ARGS', "--no-half") or ('COMMANDLINE_ARGS', "--disable-nan-check") and then save the file and try to run auto. You can always delete the --no-half or --disable-nan-check and save again if you need too. Try --no-half first.
Been awhile since I've had to do something like this but I think it is correct. Might work and if it doesn't then I'm not sure. 2.1 requires more VRAM then 1.5 does and I think 12GB is generally recommended minimum.
If you already tried this and got this error then I don't know as it works fine for me in auto1111 so I can't test it. I've had --no-half in my ARGS for like 5 months just to avoid stuff like this.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.