Qwen Image 2.1: ComfyUI editing workflow tutorial

This tutorial covers the creator's Qwen Image 2.1 Ultra workflows for ComfyUI. Installation options use a supporter installer or a separate installer for a rented RunPod GPU. The workflow files come from the creator's Patreon. The speaker describes Qwen Image 2.1 as a 7-billion-parameter model, but recommends its editing capabilities over its text-to-image results.

The multi-edit workflow accepts up to nine reference images. Examples include changing hair color, combining characters into a group portrait and transferring clothing between pictures. A prompt enhancer expands instructions, while a pixel drift fixer helps preserve an existing image. The speaker disables that fixer for new compositions and transparent PNG background removal. Bypassing the resize node retains the first image's original resolution.

The local AI editing examples also use head-swap and body-swap LoRAs. For changes confined to a specific area, the tutorial demonstrates masks and an object-removal LoRA. Its red guide marks must fit inside the mask or they can remain in the result. The speaker describes the object-moving LoRA as experimental and shows a cable complicating one edit. The final workflow extends image boundaries with an outpainting LoRA. Claims about superior quality and speed reflect the speaker's assessment of these examples.