The previous version had training errors, and this version should make the 3d style much more prominent and with less watermarks. This lora was trained with 250 different images of high quality 3D images on NoobAI-XL (NAI-XL) V-Pred-0.75S-Version.
Enjoy!
Description
FAQ
Comments (8)
What checkpoint was used for the sample images?
i put one into png info in webui forge, appears to be:
noobaiXLNAIXL_vPred075SVersion.safetensors
It might be worth retraining this lora on another trigger than "Blender".
"Blender" bleeds into the gen and puts a blender in the image. Maybe Bl3nd4r instead?
I put dick in blender
wait, no
STOP
AAAAAAAAAAAAAAAAA!!!!
Пробовал использовать illustrious чекпоинт, но вместо стиля blender, он начал выдавать аниме, подскажите в чём проблема может быть ?
It was trained specifically on the "NoobAI-XL (NAI-XL) V-Pred-0.75S-Version" checkpoint, so try using that and see if you get better results.
it says that base model is illustrious, when description says that it was trained on noob v-pred which is kinda incompatible with illustrious (because illu is eps pred)
so which one is true?
Back when I made this model, noob v-pred was categorized under illustrious on civitai. Idk what has changed since then, but when in doubt, it's probably safer to just use the model it was trained on.
Details
Files
Available On (3 platforms)
Same model published on other platforms. May have additional downloads or version variants.