4o-ghibli-at-home: locally run Ghibli-style image conversion tool

4o-ghibli-at-home is an open source AI image style conversion tool focused on converting user uploaded images into Studio Ghibli style artwork. It uses FLUX.1-Kontext-dev and DFloat11 models and runs on local devices to ensure data privacy. Users don't need to log in or rely on cloud services, the entire image processing is done locally. Currently only supports Linux, for mobile and desktop devices, with plans to support Windows in the future. the tool is suitable for creators who like the Ghibli style or anime fans, and can quickly transform ordinary photos into artistic images with soft colors and hand-painted texture. The project is licensed under the AGPLv3 license, which allows free use for personal and research purposes.

Function List

• Convert normal images to Studio Ghibli style art images with support for anime, art or custom styles.
• Provides efficient image processing using FLUX.1-Kontext-dev and DFloat11 models.
• Supports local operation, protects user privacy, and image data is not uploaded to the cloud.
• Provides an easy-to-use command line interface for technical users.
• Supports running on Linux systems and is compatible with mobile and desktop devices.
• Allows users to adjust image processing parameters through profiles to meet individual needs.
• Open source project , the code is open , allowing users to modify and optimize .

Using Help

Installation process

To use 4o-ghibli-at-home, users need to complete the installation on a Linux system. Below are the detailed steps:

1. Cloning Project Code
   Open a terminal and enter the following command to clone the project locally:

   git clone https://github.com/TheAhmadOsman/4o-ghibli-at-home.git
   cd 4o-ghibli-at-home
   

2. Creating a Virtual Environment
   Project Recommended Use uv tool to manage the Python virtual environment. If the uvPlease install it first:

   curl -LsSf https://astral.sh/uv/install.sh | sh
   

   After installation, it may be necessary to restart the terminal to ensure that the uv command takes effect. Next, create the virtual environment:

   uv venv .venv --python 3.12
   source .venv/bin/activate
   

3. Installation of dependencies
   Synchronize the Python dependencies required by the project in a virtual environment:

   uv sync
   

4. Configuring Environment Variables
   The project needs Hugging Face's access token to download the model. Copy the template file and configure it:

   cp .env_template .env
   

   Open with a text editor .env file, fill in the Hugging Face access token (generated in the Hugging Face account settings). Example:

   HUGGING_FACE_HUB_TOKEN=your_token_here
   

   If it has been passed huggingface-cli login login, there is no need to set up a token manually.

5. Starting services
   After the configuration is complete, start the project service:

   python3.12 app.py
   

   After the service is started, the project runs a local web server and background processing processes for image style conversion.

Usage

After starting the service, users can upload images via command line or web interface. The following are the specific operations:

1. Prepare the image
   Make sure there are image files (e.g. in JPG or PNG format) that need to be converted in style. Higher definition images are recommended for best results.
2. Upload images
   Once the project is running, access to the local web interface (usually the http://localhost:5000(The specific port is based on the prompt at startup). Select the image file in the interface and click Upload.
3. Choose a style
   The interface offers Ghibli style, Anime style or other customized style options. Users can choose according to their needs. The default recommended Ghibli style mimics the soft hand-drawn effect of Totoro or The Thousand and One.
4. Adjustment parameters
   exist .env file or in the web interface, the user can adjust image processing parameters such as resolution, color saturation or level of detail. Default settings can be maintained for initial use.
5. Generating images
   After uploading the image and selecting a style, click the "Generate" button. The processing time depends on the performance of the device and the size of the image, and usually takes a few seconds to a few minutes. The result is displayed on the interface and the user can download the converted image.

Featured Function Operation

• Ghibli Style Shift: The core function is to turn photos into Ghibli style images. After users upload photos of landscapes or people, the tool automatically applies soft colors, delicate textures and anime-inspired backgrounds to generate images that resemble Studio Ghibli movies.
• Privacy: All processing is done locally and images are not uploaded to the cloud, making it suitable for users who are sensitive to data privacy.
• Customized Styles: Users can extend support for other art styles, such as comics or watercolors, by modifying configuration files or code.
• Efficient processing: The FLUX.1 model is optimized for VRAM use, suitable for running on low to medium devices, and generates faster speeds.

caveat

• Make sure the device has enough storage space (model files are large and may require several GB).
• The first time you run it, you need to download the model, which may take a long time, so it is recommended to keep your network stable.
• If you encounter a model download failure, check that the Hugging Face token is correct.
• The project currently only supports Linux, Windows users will have to wait for subsequent updates.

application scenario

1. Personal artistic creation
   Animation enthusiasts can convert everyday photos, such as selfies or landscapes, into Ghibli-style artwork for social media sharing or personal collections.
2. Design material production
   Illustrators or graphic designers can use this tool to quickly generate Ghibli-style images for posters, wallpapers or concept art.
3. Education and Research
   AI researchers can use the open-source code to study image style transformation techniques and explore scenarios where the FLUX.1 model can be applied.
4. Fan Community Program
   Studio Ghibli fans can use the tool to generate images for homage works, make comics, wallpapers or peripheral designs.

QA

1. Is this tool free?
   Yes, 4o-ghibli-at-home is an open source project under the AGPLv3 license and is free for personal and research use. If used for commercial purposes, you need to comply with the license requirements and disclose the modified source code.
2. Need strong hardware support?
   Top of the line hardware is not required, but a GPU with at least 8GB of RAM and 4GB of video memory is recommended for a smooth experience.The FLUX.1 model has been optimized for VRAM use and is suitable for low to mid-range devices.
3. What image formats are supported?
   Supports common formats such as JPG, PNG. high resolution images are recommended for better conversion results.
4. How do I fix a model download failure?
   probe .env file for the correct Hugging Face token, or use the huggingface-cli login Log in. If the network is unstable, try changing networks or try again later.
5. Can you convert videos?
   The current version only supports still image conversion. Video support may be added in a future update.