Extract Product Image [Qwen & Klein]

Creates a product image for any object in the scene, be that a sitting object, held object, or worn object. It can be triggered with the prompt "make a product image for <object>"

Works best with the Base checkpoints but also works with Distilled checkpoints. If you are receiving any subject or scene bleed, try increasing the strength or adding "white background" to the prompt. Including "mannequin" in the prompt will display outfits on a mannequin rather than floating in isolation.

You can use general terms without being too specific, such as:

make a product image for her entire outfit

make a product image for her top

make a product image for his pants

make a product image for the held object

You just need to be specific enough to properly identify what you're looking for. If you get too specific, there's an observed possibility that the base model may attempt to fill in improper details (i.e. during testing, referring to monk robes as 'robes' instead of 'outfit' caused it to fill in incorrect details that would fit more for a bath robe).

California AB 2013 Training Data Disclosure

• This LoRA was fine-tuned using visual data consisting of still images collected from product listings, product reviews, and other forms of media that depict the item in a normal setting (e.g. TV stills, and still art). The training data includes copyrighted material owned by third parties. No training data was licensed or purchased. This LoRA is provided for non-commercial use only under the terms of its distribution.

• The dataset consists of 53 image pairs (106 images total). Each pair was created by locating a suitable product image and a suitable image of the object outside of a studio setting. Data was collected in 2026.

• Image data was processed through standard resizing, cropping, normalization, and labeling steps. Synthetic images and video were not included as part of the training dataset.

• This model is intended for non-commercial, experimental, and educational use. Generated outputs may reflect copyrighted visual styles or themes associated with the underlying training data. Users are responsible for ensuring compliance with applicable copyright law, other intellectual property laws, and all other applicable laws.