Improving Text-to-Image Diffusion Generation via Large Language Model

Environment

conda create -y -n diff python=3.9 cupy pkg-config compilers libjpeg-turbo opencv cudatoolkit=11.3 numba -c conda-forge
conda activate diff
pip install -r requirements.txt

Note: remember to fill in your OpenAI's API key in main.py

Imagine-Then-Verbalize

Execution example:

python main.py desc.txt -p template_prompt
python visualizer.py desc.txt -p template_prompt 

Results:

• Template-based promting: result folder and prompt folder
• CoT promting: result folder and prompt folder

Sketch-Then-Draw

Execution example:

python visualize_sketch.py

Results: result folder and prompt folder