QuadraMuse — Four artistic voices, one canvas.

What it does

Turn a single photo into a single composite artwork by reimagining different parts of the image in different art styles (cubist, modern abstract, impressionist, surrealist) and blending those parts together into one unified piece.

Quickstart (PowerShell)

1. Create and activate a virtual environment

python -m venv .venv
.\.venv\Scripts\Activate.ps1

2. Install dependencies

pip install -r requirements.txt

3. Set your Hugging Face API key (recommended) — do NOT hardcode keys in the script

$env:HF_API_KEY = "your_hf_api_key_here"

4. Run the tool

python .\try.py input.jpg output.png --max_calls 4

Options

• --max_calls (0-4): limit how many regions are sent to the remote stylization API. Default: 4.
• --randomize: choose regions randomly when --max_calls is less than the number of non-empty regions.
• --seed: optional deterministic seed for mask generation and random selection.

Notes

• Each stylized region makes an API request to the image-to-image model; watch your quota and costs.
• The script currently uses huggingface_hub.InferenceClient for remote image-to-image calls. If you want to run models locally, install torch and diffusers and adapt the code.
• The script intentionally avoids caching so each run reflects the current input and parameters.

License

Use this code as you like for experimentation and prototyping.