Trained using only 7 similar images, so it's somewhat rigid with lighting. (Fixed in V2)
Tags:
Appearance:
2dmita, 1girl, solo, blue eyes, blue hair, short twintails, medium hair, red bow, hair bow, hairclip,Outfit:
red choker, ribbon choker, red shirt, short over long sleeves, striped sleeves, pleated skirt, shirt tucked in, miniskirt, blue skirt, red socks, kneehighs, Add to negative:
bloom, glowing, Description
FAQ
Comments (2)
Friend, could you share how you train with only 7 images? Or some kind of guide you usually follow?
In short Illustrious doesn't need much training data, you can get great LORA with 10 images, just make sure your dataset is properly tagged, and preferably has different angles. Prodigy optimizer seems to work great for small dataset.
(NoobAI might be even better with small datasets, I haven't done enough testing)
I trained this LORA for NoobAI using 10 images - https://civitai.com/models/1059679/luminous-valentine-tensura-tensei-shitara-slime-datta-ken
If you want i can give you training data and training settings I used for training LORA of some random background character using only 5 images with very good results - https://imgsli.com/MzQwMzUy
DM me if you need it.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.