IP-Clone

I use IP-Adapter FaceID along with Stable Diffusion model to transform the face from an uploaded image into a new image based on textual prompts.

Installation

To set up the application, follow these steps:

1. Clone the repository and enter its directory:

git clone https://github.com/Ibtisam-Mohammad/IP-Clone.git
cd IP-Clone

2. Clone the IP-Adapter repository inside the IP-Clone directory:

git clone https://github.com/tencent-ailab/IP-Adapter.git

3. Install the required dependencies:

pip install -r requirements.txt

4. Download the necessary model file:

wget https://huggingface.co/h94/IP-Adapter-FaceID/resolve/main/ip-adapter-faceid_sd15.bin

Usage

To generate an image from a face in an uploaded image, run the generation script with the desired prompt. For example, to generate an image based on the prompt "A man in a library":

python queue_gen.py --base_path "/content/IP-Clone/" \
                    --image_name img.jpeg \
                    --prompt "A man in a library"

Input Image

Generated Image

Note:

• We can use a better diffusion model like SDXL (or faster one like LCM) or another version of IP-Adapter , but I stuck to the basic version. Although it can be easily changed.
• If AWS isn't available, passing arguments for local testing, should simulate the generation process, otherwise we have to setup the AWS account and pass the related arguments.
• Generated result is saved in the same directory as the uploaded image.
• Exception handling could be made more robust, but I set it to the basic for MVP.
• Reference code: https://huggingface.co/h94/IP-Adapter-FaceID