anime-shoujo-sparkle

Pastel Romantic SDXL LoRA for Shoujo Art

What it does

This LoRA enhances the classic shoujo anime aesthetic with a soft pastel palette, large expressive eyes with highlights, flowing hair, floral accents, and sparkling details. It creates a dreamy romantic atmosphere perfect for depicting young girls in various scenarios.

How to use

Apply the trigger word 'sh0ujo' followed by your prompt to invoke this style. A suggested strength of 0.85 works well with most samplers; however, you may find it effective at strengths between 0.6 and 1.0.

Best for

• Depicting young girls in romantic or whimsical settings
• Illustrating scenes from shoujo manga or anime
• Adding floral and sparkly accents to portraits
• Creating dreamy backgrounds with gentle blushes
• Enhancing the overall aesthetic of a scene with decorative bordering

Tips

Use 'sh0ujo' as early in your prompt as possible for best results. Consider adjusting strength based on how dominant you want this style to be.

Example Prompts

sh0ujo portrait of a schoolgirl with flowing hair, soft pastel colors, gentle blush, large eyes with sparkles and highlights

sh0ujo an enchanted garden scene with delicate floral accents, sparkling light effects, romantic atmosphere, decorative bordering

sh0ujo close-up of a pink carnation in full bloom with subtle sparkle details, set against a pastel background, dreamy lighting

sh0ujo a ballet dancer mid-pirouette, graceful lines and movement captured, large expressive eyes, flowing hair, soft lighting

sh0ujo quiet afternoon at the library, warm glow from windows, gentle mood with hints of romance, decorative bordering around image

sh0ujo abstract swirls of pastel colors blending into each other, subtle sparkle accents, dreamy and romantic atmosphere

Trigger word

Activate with sh0ujo. Recommended strength: 0.9.

Showcase

Usage with diffusers

from diffusers import AutoPipelineForText2Image
import torch

pipe = AutoPipelineForText2Image.from_pretrained("stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16).to("cuda")
pipe.load_lora_weights("strkyyy/anime-shoujo-sparkle")
image = pipe("sh0ujo, your prompt", num_inference_steps=30).images[0]