Flux2klein enhancer
New experimental node lets users target specific image areas for editing while preserving original structure.
Developer Capitan01R has launched a BETA experimental update to the Flux2Klein Enhancer, a node pack for the popular ComfyUI workflow system. The key addition is the 'FLUX.2 Klein Mask Ref Controller,' a novel approach to regional conditioning for the FLUX.2 Klein model. Unlike traditional inpainting, this node uses masks to spatially control reference latents directly within the conditioning stream, allowing for precise, structure-preserving edits to specific areas of an image.
The node enables users to apply prompts to masked regions while maintaining the original subject's structural integrity, with unmasked areas becoming fully available for new content generation. Early testing demonstrates the tool's ability to handle complex scenarios, such as isolating and modifying one character in a three-person photo or implementing smooth zoom effects. The developer reports promising results in preserving fine subject details while permitting significant background and environmental changes, all without the model hallucinating new structures. The tool is available via GitHub, with example workflows provided for drag-and-drop implementation.
- New 'FLUX.2 Klein Mask Ref Controller' node enables mask-guided regional conditioning, not inpainting
- Uses masks to spatially control reference latents in the conditioning stream for precise edits
- Early tests show successful subject isolation in multi-character images and structure-preserving background changes
Why It Matters
Provides artists and developers with finer control over AI image generation, enabling complex edits without losing structural integrity.