diffusion_flow_inversion is a versatile repository built using the diffusers library, designed to work with various diffusion models such as SD1.5, SD2.0, SD2.1, SD3.0, Flux, and other flow-based models. This repository offers powerful tools for visualizing the transformation process between images and noise, allowing users to observe and analyze the differences in the image-to-noise-to-image cycle and noise-to-image-to-noise cycle.
- Support for Multiple Diffusion Models: Seamlessly integrates with SD1.5, SD2.0, SD2.1, SD3.0, Flux, and other popular diffusion and flow models.
- Visualization of Transformation Processes: Provides visualization capabilities to study the transitions from image to noise, and from noise to image, as well as the differences across these stages.
- Customizable Time Schedules: Allows users to define custom time schedules to experiment with various parameters, making it highly effective for research on image editing and reversible image reconstruction.
This repository is particularly useful for those looking to explore the intricacies of image generation and manipulation, and is a powerful tool for researchers in fields such as computer vision, generative models, and image restoration.
- Clone this repository:
git clone https://github.com/yourusername/diffusion_flow_inversion.git pip install -r requirements.txt