Detextr

Detextr is a tool for repainting styled text from images using Stable Diffusion with canny edge Control-Net and font style transfer.

Exhibition

basic usage for demo

1. for the first time, install the required packages by running the following command in the terminal:

pip install -r requirements.txt

2. (optional) we also provide a pre-trained Lora model for you to use. You can download it from here and place it in the weights directory.

3. download pretrained models from TextEffects and the example image from exampe_image

4. unzip the downloaded FET-GAN pre-trained model and place it in the weights directory.

5. unzip the downloaded FET-GAN pre-trained model and place it in the workspace as the image_repos directory.

6. now you should have a directory structure like this:

.
├── configs
├── fft
├── image_repos
├── models
├── networks
├── process
├── utils
└── weights

6. run the following command in the terminal( for the first time, pre-trained diffusion model would be downloaded, which may take tens of minutes):

python detextr.py 1 1 [text]

such as:

python detextr.py 1 1 A

7. the result will be saved as result.png in the workspace.

Related works and Thanks

• Diffuser

• FET-GAN