If you have questions, suggestion, problem, or just like the model feel free to comment.
LoRA Training Info I trained this LoRA using 157 images of Lady Silver Lily with 6,000 steps.
The dataset contains no AI-generated content.
Important Warning Using heavy quality tags such as (masterpiece), (best quality), (ultra-detailed), amazing quality, very aesthetic, illustration, perfect composition, intricate details, absurdres, incredibly absurdres
will reduce the LoRA’s character fidelity. The images may look more beautiful overall, but the character will no longer resemble Raviel accurately.
Essential Prompts
Character: white eyelashes is optional
1girl, Raviel, white eyelashes, long wavy silver hair, striking red eyes
1girl, Raviel, long wavy silver hair, striking red eyes
Outfits:
School Uniform: white ruffled blouse, white jacket with black trim, decorative brooch, black pleated skirt, gold trim, school uniform
Nightgown / Pajamas: white gown, nightgown, ruffled blouse with delicate lace details, transparent clothing, see-through clothes
Blue Dress: off-shoulder evening dress, blue floral gown, embroidered patterns, high slit dress, fitted waist, ornate bodice, lily decorations, gemstone ornaments, long gloves, blue gloves
You can copy Raviel prompts from other LoRAs and test them as well. However, I did not train on those prompts or use standard booru-style tagging in my dataset.
Example of a prompt I used in dataset: “A person with long, flowing white hair and striking red eyes, wearing an ornate, deep red Victorian-style gown with lace and floral embroidery, seated against an intricate golden circular background. The character has a pale complexion, rosy cheeks, and a subtle smirk, with one hand resting near their chest. The gown features a high collar with lace detailing and a fitted bodice with ruffled sleeves. The background is composed of elaborate gold filigree patterns, suggesting a luxurious or regal setting.”
You can also use regular booru-style prompts — they work great too, just like the images I posted.
I basically copy jose prompt and test it. idk why he's using score prompt on Illustrious.. whatever.
his lora is great i have been using a lot. thank you for making a lot of great loras.
If you use non finetuned model like IlustriousXL2.0 (i trained on this)
She will look exactly like she comes straight from manwha.
