Differential Diffusion is a platform for image generation and editing that allows modification of images based on text prompts and a specified map that defines the variation in each region. It provides customized variation control for each pixel or image area, opening doors to new editing capabilities such as controlling the extent of modification for individual objects or introducing gradient spatial changes. In addition, the platform demonstrates the effectiveness of this framework in the field of image completion, particularly fine-tuning the surrounding areas when seamlessly blending new content. It also offers new tools to explore the effects of different variation amounts. This framework operates solely during inference and does not require model training or fine-tuning. It showcases its integration with the most advanced open-source models and has been validated through quantitative, qualitative comparison, and user studies.