ArtificialGuyBR

Home / Blog / Training LoRAs

Studio Ghibli LoRA for SDXL: Generate Ghibli-Style Art

Use the StudioGhibli.Redmond LoRA with SDXL to create Studio Ghibli style illustrations. Trigger words, settings, and examples.

2 sources cited Training LoRAs

On this page

  1. What This Model Does
  2. Trigger Words & Prompting
  3. Recommended Settings
  4. Usage in Popular UIs
  5. License & Attribution
  6. Common Errors and Fixes
  7. FAQ
  8. Sources

Studio Ghibli LoRA for SDXL: Generate Ghibli-Style Illustrations

TL;DR

What This Model Does

StudioGhibli.Redmond V2 is a LoRA (Low-Rank Adaptation) trained on top of stabilityai/stable-diffusion-xl-base-1.0. It adapts the base model to generate images that mimic the distinctive visual language of Studio Ghibli films: soft watercolor backgrounds, expressive character designs, whimsical creatures, and that signature blend of fantasy and everyday life.

The model was trained on a large curated dataset with GPU compute sponsored by Redmond.AI. It's distributed as a SafeTensors file (~104 MB) and loads directly in any SDXL-compatible interface (ComfyUI, Automatic1111, diffusers, InvokeAI).

Studio Ghibli style portrait of a boy

Trigger Words & Prompting

Two tokens activate the style:

Include both in your prompt. Example:

Portrait of a boy, stunning, cute, Studio Ghibli, StdGBRedmAF

Recommended negative prompt (from the model card):

bad art, ugly, text, watermark, duplicated, deformed

Parameter | Value

|-----------|-------

Base Model | SDXL 1.0 (stabilityai/stable-diffusion-xl-base-1.0) Sampler | DPM++ 2M Karras / Euler a Steps | 20–30 CFG Scale | 5–7 LoRA Weight | 0.8–1.0 Resolution | 1024×1024 (SDXL native) The author notes the LoRA also excels at coloring book style line art — reduce CFG to 3–4 and add "line art, coloring book, black and white" to prompts for that variant.

Cute yellow fish in water, Studio Ghibli style

ComfyUI: Load via Load LoRA node, connect to model and clip, set strength 0.8–1.0.

Automatic1111: Place .safetensors in models/Lora, add <lora:StudioGhibli.Redmond-V2:0.9> to prompt.

diffusers (Python):

from diffusers import StableDiffusionXLPipeline
import torch

pipe = StableDiffusionXLPipeline.from_pretrained(
    "stabilityai/stable-diffusion-xl-base-1.0",
    torch_dtype=torch.float16,
    variant="fp16"
).to("cuda")

pipe.load_lora_weights("artificialguybr/StudioGhibli.Redmond-V2")

License & Attribution

Non-commercial, fan-made creation for entertainment purposes. All character rights belong to Studio Ghibli / respective owners. Contact [email protected] for concerns. The author accepts support via Patreon, Ko-fi, and Buy Me a Coffee.

Common Errors and Fixes

Error | Cause | Fix

|-------|-------|-----

Style not appearing | Missing trigger words | Include both Studio Ghibli and StdGBRedmAF Overcooked / artifacts | LoRA weight too high | Reduce to 0.7–0.8 Wrong base model | Using SD 1.5 or SD 2.1 | Use SDXL 1.0 base (stabilityai/stable-diffusion-xl-base-1.0) Color palette off | CFG too high | Lower CFG to 5–6

FAQ

Q: Does this work with SD 1.5 or SD 2.1? A: No. This LoRA targets SDXL 1.0. The Civitai page lists a separate v2.1 variant for SD 2.1.

Q: Can I use this commercially? A: The author states it's non-commercial and fan-made. Check with the author for commercial licensing.

Q: What's the file size? A: ~104 MB (SafeTensors).

Q: Who created this? A: ArtificialGuyBR, with GPU sponsorship from Redmond.AI.

Sources