Fast Aesthetic Image Generation for Paper-Cutting Style

This paper proposes a efficient novel workflow based on stable diffusion named SAP, in which ControlNet is incorporated to regulate the structural consistency of images throughout the transformation process. Furthermore, a LoRA model specifically trained for Chinese paper cutting is applied to achieve stable and effective style transfer. Extensive image and numerical data demonstrate that the SCP method effectively enhances the aesthetic appeal of images in the Chinese paper-cutting style while preserving the original image structure and leveraging the powerful text-to-image capabilities of diffusion models. Specifically, with this method, the machine-evaluated aesthetic score improves by approximately 10.84%.