Previously, when I was watching videos, I thought Lilith was very beautiful. But I couldn't find the corresponding Lora. I'm not sure if it's because I didn't search for it correctly. So I followed the tutorial and trained one myself.

The photo editing tool is SDforge. The others are set to the default values.

Strength recommendation

<lora:lilith_SDXL_V2:0.7>

Basic characteristics

Ponytail hairstyle

"llth,1girl,solo,blue eyes,blonde hair,braid,ponytail,crown braid,long hair,earrings,jewelry,"

2. Long loose hair hairstyle

"llth,1girl,solo,blue eyes,blonde hair,long hair,earrings,jewelry,"

Clothes

"collarbone,bare shoulders,white dress,white gloves,elbow gloves,black thighhighs,black high heels,"

PS:

1. Since the images in the training set are screenshots from the movie scenes, the colors of the original images are greatly affected by the environment. Even if they are adjusted for lighting in Photoshop, the effect is still not good. The original images are likely to be backlit.

2. The black veil on the chest. For some angles in the training set, there are no corresponding images, which means that some angles may not be able to be generated when creating the original image.

3. There is an image with a hat. The available training images are limited, and this basically indicates failure.

4. The transparent shawl. The images processed might not have been handled properly, and it can only be generated normally in certain environments. The trigger word "capelet, see-through" is not recommended.

~~~~~~

When I first trained the model, I was still inexperienced when selecting the training set and processing the dataset. If you have any special techniques, please feel free to ask me.