MaterialShift_Klein_V1.0

MaterialShift_Klein_3D Material Transfer is a LoRA model based on Flux.2 Klein, designed to solve the common limitation of surface-only material replacement.

Instead of simply applying a flat texture, this model transfers the material characteristics from a reference fabric/material image onto the target object in Image 1 while preserving the original 3D structure, silhouette, lighting logic, seams, proportions, and pose.

The training set focuses primarily on clothing materials, so the model is especially stable on structured fashion categories such as jackets, coats, pants, and outerwear. It also supports cross-category generalization for objects with clear material properties, including furniture, soft furnishings, and other material-sensitive products.

Key Features

True 3D Material Transfer

Transfers not only the visual texture but also the perceived thickness, wrinkle response, reflectivity, softness, stiffness, and shadow behavior of the material.

Clothing-Oriented Training

Trained with a large amount of garment material data, making results more natural for clothing structure, body tension, seams, folds, and garment drape.

Structure-Preserving

Keeps the original composition, silhouette, stitching, proportions, and pose of Image 1 while changing only the material appearance.

Cross-Category Support

Beyond clothing, the model can be applied to furniture, soft products, and other objects with well-defined material attributes.

Trigger Prompt

Change the material of the [target category] in Image 1 to match the material in Image 2.

Recommended target categories should be specific, for example:

jacket
coat
sofa
chair
pants

Example Usage

Use a clear target image as Image 1, then provide a material or fabric reference image. In the prompt, explicitly name the object whose material should be changed.

Example:

Change the material of the jacket in Image 1 to match the material in Image 2.

Recommended Use Cases

• Fashion material visualization
• Jacket, coat, pants, and outerwear material replacement
• Fabric-to-garment transfer
• Product design material exploration
• Furniture and soft furnishing material transfer
• E-commerce visual material variants

Limitations

• Best results are expected when the target object has a clear 3D form and visible material behavior.
• Clothing categories are the most stable because they are the primary focus of the training data.
• Very flat objects or ambiguous target regions may produce weaker material transfer.
• Prompts should clearly identify the target object.

Files

• MaterialShift_Klein_V1.0.safetensors: LoRA weights
• images/: example illustrations

Gallery