Hey there!
I’m new to Loras and I need a bit help with creating one.
I’m trying to create my own Lora on tensor art website. However, when my trainings are done my images with my Lora in stable diffusion (forge UI) are very dim (see example below with strength 1.0).
At first, I thought that the source images may cause the problem. So I created a new model using new photos but the problem wasn’t fixed.
I have 20 HQ images with a persistent face. As a base model for training I’ve selected pikonRealismV2 (the one I use in Forge UI) and I set 20 repeats and 10 epochs. The rest settings are left as they are.
P.S: lowering the strength of my Lora (0.5 and below) makes the image clearer and richer with colors, but face details (face from Lora) are washing away. Setting Lora to 1.0 makes the image dim, poor with details and colors, but makes the face consistent (with my Lora)
What may cause the issue? Any ideas?
Thanks!
