sd15-storybook-anime-lora

Storybook Anime LoRA for Stable Diffusion 1.5

<Gallery />

Model description

This is a LoRA adapter for stable-diffusion-v1-5/stable-diffusion-v1-5 tuned toward soft storybook anime illustrations, warm hand-painted backgrounds, gentle cel-shaded characters, cozy interiors, gardens, fields, cottages, bakeries, and simple children's-book style scenes.

The repository contains the LoRA adapter only. It does not include or redistribute the base model.

Trigger words

Use:

storybook anime illustration

Recommended settings

• LoRA weight: 0.75-0.95
• Starting weight: 0.85
• Steps: 25-35
• CFG scale: 6.5-7.5
• Sampler: DPM++ 2M Karras or Euler a
• Resolution: 512 x 512 for native SD 1.5, or up to 768 on one side with hires/upscale if your workflow supports it

Prompting guidance

This LoRA tends to respond best when prompts keep the style phrase near the front and end with a few consistent style anchors:

storybook anime illustration of [subject and action], [simple setting clue], warm hand-painted background, soft cel-shaded characters, whimsical children's book anime style

Useful anchors include warm hand-painted background, soft cel-shaded characters, cozy scene design, gentle atmosphere, storybook anime illustration, and whimsical children's book anime style.

Training data

Trained on 70 GPT-generated storybook anime style images with matching text captions.

Training settings

• Base model: stable-diffusion-v1-5/stable-diffusion-v1-5
• Training platform: Modal
• GPU: L4
• Steps: 2400
• Resolution: 512 x 512 with buckets
• Batch size: 4
• Network rank: 32
• Network alpha: 16
• Network dropout: 0.05
• Optimizer: AdamW8bit
• U-Net LR: 1e-4
• Text encoder LR: 5e-5
• LR scheduler: cosine

File

storybook_anime_lora.safetensors

SHA256:

0304bb42819790cf63f4125ac70cee60fda682cc5aa4bfb512bb9e3b1746e9b0

License

This LoRA follows the base model license: CreativeML Open RAIL-M.