Upscale image

This section contains examples of image upscaling using various models and workflows.

Each example demonstrates the results of different upscaling approaches applied to test images.

Test images:

Real-ESRGAN (Real-ESRGAN: Towards Practical Real-World Super-Resolution) is an advanced Deep Learning model used for Super-Resolution (SR), which is the process of intelligently increasing an image’s resolution.

It is an advanced finetune based on the ESRGAN/RealESRGAN deep learning architecture, optimized for generating realistic textures and details.

This model is designed to add details but generally does not perform strong denoising or deblurring. It focuses primarily on realistic texture synthesis.

4x-UltraSharp is a popular custom Super-Resolution (SR) model frequently used in ComfyUI and other Stable Diffusion environments. It is a highly effective AI model designed for image upscaling.