Type | |
Stats | 106 2 |
Reviews | (15) |
Published | Jan 24, 2025 |
Base Model | |
Training | Epochs: 10 |
Usage Tips | Clip Skip: 2 Strength: 0.8 |
Trigger Words | planar dimly lit brightly lit stylized planar greyscale head only torso only planar male head planar female head planar male body +1 more |
Hash | AutoV2 EDAED3A53F |
This is an experimental lora I made to help me better understand planar changes in the face for my 3d sculpts. I'm posting this here for any other 3d artists in the A.I. community that might find this useful. The dataset only contained stylized zbrush sculpts of people but I have found through testing that this lora can "planarize" things well outside of the trained data.
Prompt formatting:
(planar, greyscale) {your prompt}, from front perspective, dimly lit, black background,
You can be a little looser with the prompt, but be sure to include planar before any subject tags. This should increase the effect of the "planarization" of your subject.
Other tags:
dimly lit -> smooth shadows and bright highlights on areas of transition
brightly lit -> bright images, very soft shadows
stylized planar -> gives it a bit more of a Pixar 3d look
head only -> should isolate the head (not totally effective in my testing)
planar head, planar torso, planar hair, etc -> This seems to better target the planarization, I'd recommend a pretty heavy use of the "planar" tag
Current drawbacks:
Img2img/adetailer problems. This lora really struggles with img2img and inpainting.
I'd recommend turning down the weight of the lora to .65 as well as the denoising strength below 50
2d models: While mixing this lora with 2d character loras it seems to perform fairly well. However, when using this Lora in a 2d base model, it doesn't typically give the intended effect. More experimentation may be needed as I do not personally use many 2d models myself.
Limited dataset: This Lora was only trained on a very limited dataset as it was only intended for personal use. With this in mind the variety and versatility may be a bit lacking.