[Concept version]
I noticed that some checkpoints have trouble with glasses, like rims fading into nothingness and etc. So this is my first LoRA that I botched up in 1 hour and trained for 20 minutes on like 15 images. Nothing groundbreaking, but this first concept version works (For Aviators).
Trained on: SXZ-Luma model
Recommended model: DarkSushiMix
Recommended weight: 0.4 - 0.8
Useful trigger words: aviators, sunglasses, reflection, tinted
Recommendations: Use low GFG scale (5.5), use more Steps (40-50), trained only on 512x512 images, so higher resolution is a bit wonky at higher weights.
Enjoy prompting!
Description
Added
FAQ
Comments (3)
does first concept version mean you plan to support more types of glasses in future versions of this lora? would be awesome! plz?
Yes, indeed. The first version was more like a proof of concept to see if I am even able to make LoRA. No inclusive dataset, no precise descriptions, just quick training... to see if it works. For the little effort that I did I think I got some amazing results. Now that I know the process and I played a little bit with it, I can make bigger and more precise dataset on variety of styles and glasses types. I will give it more time to train too. Stay tuned!
great idea, awesome!!
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.