Virtual Try-on Workflow with TryOnDiffusion

TryOnDiffusion, a virtual try-on AI method which superimposes any clothing onto a person’s image, considering body shape and pose. It offers a detailed, realistic virtual try-on experience. Learn how you can build virtual try-on workflow on Pixelflow.

Virtual Try-on Workflow with TryOnDiffusion

Given separate images of a person and an article of clothing, TryOnDiffusion generates a realistic image of how the clothing would look on that person. TryOnDiffusion, a virtual try-on AI technique can realistically superimpose clothing onto a person's image despite variations in body shape and pose. By warping the garment image and blending it seamlessly with the person's image, TryOnDiffusion offers a more accurate and detailed virtual try-on experience, potentially transforming online shopping. This has a huge potential in the ecommerce industry.

Virtual Try-on with TryOnDiffusion

How does TryonDiffusion work?

TryOnDiffusion uses a system called Parallel-UNet. It is designed to address two challenges in virtual try-on. One is preserving the garment or clothing details and the other one is adapting to body pose and shape changes. The parallel-Unet has two parts: Person-UNet and Garment-UNet.  It combines two UNets, each specializing in a different task. One UNet, the "person-UNet," analyzes a picture of a person with some added noise to capture detail. The other, the "garment-UNet," focuses on the image of the clothing that needs be to overlayed on the person. To make the clothes fit perfectly, TryOnDiffusion uses a technique called "implicit warping." This lets the garment-UNet adjust the clothing's shape on the fly, while the person-UNet seamlessly blends it onto the person's image. To keep things efficient, the garment-UNet stops processing at a certain point, focusing its efforts on the most crucial stages. This two-UNet teamwork with implicit warping is what allows TryOnDiffusion to create such realistic virtual try-on experiences.

How to Use TryOnDiffusion?

The TryOnDiffusion model in PixelFlow can handle upper body, lower body, and full body virtual try-ons. This means you can see how different types of clothing items, whether it’s a shirt, pants, or a full outfit, would look on someone.

Upper Body Virtual Try-on

This involves overlaying clothing items such as t-shirts, tank tops, shirts, jackets, etc., onto the model images. You can choose the upper body as a category in the TryOnDiffusion model for upper body virtual try-on.

Upper body virtual try-on

Lower Body Virtual  Try-on

Similar to the upper body try-on, this involves overlaying lower body clothing items like pants, trousers, skirts, shorts, etc., onto the model images. You can choose the lower body as a category in the TryOnDiffusion model for the lower bosy virtual try-on.

Lower body virtual try-on

Full Body Virtual Try-on

This is an extension of the upper and lower body virtual try-ons. It involves overlaying a complete dress onto the model images. This provides a more comprehensive view, allowing how a complete outfit would look on them. You can choose the full body as a category in the TryOnDiffusion model for the full body virtual try-on.

Full body virtual try-on

We have built a ready to use workflow template on Pixelflow using TryOnDiffusion model. You can access it by simply copying it and trying with your set of clothing.

Virtual Try-on workflow using TryOnDiffusion on Pixelflow

Get Virtual Try-on workflow using TryOnDiffusion on Pixelflow for free.