Generate unique-looking art using style-transfer AIs along with random images.
This project is created with:
- opencv-contrib-python: 4.5.1.48
- Numpy: 1.18.5
- Pixabay's API
- Deepai's Style Transfer API
For whenever you need a constant stream of unique-looking images, this project has you covered. It gets a random image from a Pixabay query, overlays a number of random 'stickers' (any file in ./Overlays/), fills transparency with an image from ./Underlays/ before completely changing the content using the style-transfer AI
All included images are from [Pixabay](https://pixabay.com) and are licensed under [Pixabay's license](https://pixabay.com/service/terms/#license)
To run this project, simply download it and install the requirements with pip3 install -r requirements.txt.
Then, specify your API key for pixabay in ./pixabay_key.txt
Modifying the contents of ./Overlays/, ./Styles/, and ./Underlays/ is recommended, but the program works out-of-the-box.
Simply running the program with python3: python3 generate.py allows for the user to input a prompt.
For implementation in your own projects, the apply_style(image, style, is_url, verbose=True, retries=10, use_proxies=True, set_proxy='') function uses Deepai's Style Transfer API to apply a style to your own opencv images (or image urls).