Okay, trying to get an AI to realistically put new clothes on a person is usually a total mess. You end up with floating shirts or results that look like a bad copy-paste job.
But I just saw a post from Tianyu Xu that completely cracks the code, and I had to share it. He’s using a feature called Nano Banana, and it’s a game-changer for virtual try-ons.
✨ The Simple Trick Tianyu Discovered
He found that the magic lies in a super simple prompt structure. You just give the AI a character, a product photo, and tell it:
“Take an element from [Image 1] and add it to [Image 2].”
That’s the core of it!
✍️ His Pro-Level Prompt
To avoid that ‘pasted on’ look, Tianyu shared his secret for getting natural-looking results. He adds more detail about context and lighting.
Check out his example:
“Put the outfit from [Image 1] on the man from [Image 2], placing him in the [context], ensure his face is clearly visible and adjust the lighting on his face.“
Adding those extra details is what makes it blend perfectly.
🚀 Tianyu Xu’s Full Workflow
He even laid out his entire process for creating awesome product videos:
- Generate Outfits: He used a GPT model to create a set of cool outfits.
- Virtual Try-On: He then used Nano Banana to put those outfits on his own picture.
- Animate: Finally, he used Kling v2.1 to turn the images into super smooth videos.
He mentioned this is going to be huge for e-commerce and design, and that it works great on tools like Weavy and OpenArt where you have more control.
Huge props to Tianyu Xu for this awesome discovery and for sharing such a practical workflow. Go check out his original post to see the video he made!