FLUX Virtual Try-On
Upload a person photo and a garment image. Generate a virtual try-on result with the garment applied to the person while preserving face and pose.
例を試す
このワークフローについて
This ComfyUI workflow performs virtual try-on by taking two inputs—a person photo and a garment image—and compositing them into a realistic outfit preview while preserving the subject’s face and pose. The core is the FluxVTONode (FLUX Virtual Try-On model), which receives the Person (LoadImage) and Garment (LoadImage) inputs, then outputs a single try-on result that keeps identity features from the person image and transfers the garment’s texture and shape.
Around the model node, utility nodes make the process practical and reviewable. ImageCompare lets you place the original person image alongside the try-on result for instant before/after inspection, and ImageStitch can assemble a simple contact sheet (e.g., garment, person, and result in one panel). Finally, SaveImage writes the synthesized try-on to disk. The included MarkdownNote summarizes quick-start guidance so you can load images, run the queue, and iterate quickly with minimal configuration.
よくある質問
For the person: a front-facing photo with the torso clearly visible, good lighting, and minimal occlusions (crossed arms, bags, or hair covering the neckline). For the garment: a front-facing product image on a simple background with the item fully visible. Clean, well-framed inputs help the FluxVTONode align and transfer details more accurately.
No. The FluxVTONode is designed to preserve the subject’s face and overall pose from the person image. If you notice artifacts near the neckline or face, try a person photo with fewer obstructions and a garment image that clearly shows the collar/shoulder area.
The workflow works best with tops, tees, shirts, jackets, dresses, and similar front-view garments. Highly complex items (e.g., scarves, very loose sleeves, layers) or extreme poses can be challenging and may produce distortions. Clear, frontal views improve results.
Use ImageCompare to view the original person image next to the try-on output for quick checks. ImageStitch can create a single panel with garment, person, and result for sharing. Adjust the SaveImage node’s directory/name to export the final image.