Fashion-VDM is a breakthrough virtual try-on technology jointly launched by Google and the University of Washington.

It uses the Video Diffusion Model (VDM) to generate high-quality virtual try-on videos, which can truly display the effect of clothing under different movements, perfectly maintain the identity and movement details of the characters, and significantly improve the realism and detail performance of clothing display.

Feature Highlights

• The identity and movement of the character are consistent: Fashion-VDM can generate a virtual try-on effect with just an image of clothing and a video of the character. No matter how the characters' expressions and body postures change, the generated videos maintain high consistency and show a realistic wearing experience.
• Detail-rich and natural: Thanks to the advanced video diffusion model, Fashion-VDM is able to accurately capture the complex details of clothing, such as texture, sheen and folds, making virtual try-on videos more natural, clothing More fitting to the character.
• Smooth time consistency: In the time dimension, Fashion-VDM shows excellent consistency. The characters' movements are smooth and natural, with no frame skipping or discontinuity, ensuring the continuity of the try-on video.
• Real dynamic interactive display: Compared with traditional image virtual try-on, video try-on is more vivid. For example, details such as the folds and swings of clothing caused by the movement of characters are accurately captured, making the interaction between clothing and characters more realistic.

Comparison with traditional virtual try-on technology

• Detail capture: Fashion-VDM greatly improves the presentation of clothing details, such as the delicate expression of texture and changes in gloss. Compared with image generation methods, the effect of clothing in the video is more natural.
• Action Fluency: Traditional methods are prone to frame skipping and discontinuity in videos, while Fashion-VDM has significantly improved coherence and fluency, including details of characters’ movements and dynamic changes in clothing. Be consistent.
• Dynamic performance: Compared with static image virtual try-on, Fashion-VDM's video try-on can more realistically show how clothes change with the movements of people, providing an immersive try-on experience .

Application scenarios

Fashion-VDM has a wide range of application prospects, especially in the following fields:

• Online clothing retail: Consumers can more intuitively see the actual wearing effect of clothes, feel the texture and fitting effect of clothes, and reduce the risk of purchasing inappropriate clothes.
• Virtual fashion display: Fashion brands can show the design details of clothing through virtual try-on videos, allowing customers to better understand the material and style of clothing.
• Personalized recommendations: Provide users with personalized clothing suggestions to help users quickly find clothing combinations that suit them.

Fashion-VDM has created a more realistic and smooth try-on method in the field of virtual try-on, which will bring a new perspective to online shopping experience and fashion display!

Project Link: https://johannakarras.github.io/Fashion-VDM