https://github.com/igoldshm/ghibli-style-transfer-lora
A lightweight style transfer model using LoRA fine-tuning on Stable Diffusion to convert real-world photos into Studio Ghibli-inspired art.
https://github.com/igoldshm/ghibli-style-transfer-lora
ghibli image-generation lora stable-diffusion
Last synced: 3 months ago
JSON representation
A lightweight style transfer model using LoRA fine-tuning on Stable Diffusion to convert real-world photos into Studio Ghibli-inspired art.
- Host: GitHub
- URL: https://github.com/igoldshm/ghibli-style-transfer-lora
- Owner: igoldshm
- Created: 2025-04-16T06:19:02.000Z (6 months ago)
- Default Branch: main
- Last Pushed: 2025-05-23T05:18:57.000Z (5 months ago)
- Last Synced: 2025-05-23T06:46:41.983Z (5 months ago)
- Topics: ghibli, image-generation, lora, stable-diffusion
- Homepage:
- Size: 5.86 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# ghibli-lora-generator
# Ghibli Style Transfer with LoRAConvert real-world photos into beautiful Studio Ghibli-inspired art using lightweight LoRA fine-tuning on Stable DiffusionXL (SDXL).
## Project Overview
This project demonstrates an efficient style transfer pipeline where input photographs are transformed into Ghibli-style images. The model was fine-tuned using the [LoRA](https://arxiv.org/abs/2106.09685) (Low-Rank Adaptation) method on top of the `runwayml/stable-diffusion-v1-5` base model.
---
## Features
- **Photo-to-Ghibli**: Stylizes any input photo into a soft, dreamlike Ghibli aesthetic
- **LoRA Training**: Lightweight training (~100MB) instead of full model fine-tuning
- **Built on Stable Diffusion v1.5**
- Supports inference using HuggingFace `diffusers` or AUTOMATIC1111 Web UI
- Easy to adapt to other styles (just swap dataset)---
## Project Structure