Doppl turns uploaded outfit photos into animated try-on videos

What Doppl actually accepts and returns

Doppl is a Google Labs app, launched June 26, 2025 on iOS and Android in the U.S. The core loop is simple to describe: you upload a photo of an outfit, and the app shows that outfit on a digital, animated version of you.

Two details separate it from a standard catalog try-on. First, the input isn't limited to a store's product feed. The announcement is explicit that the source can be a friend's outfit, something at a local thrift shop, or a look featured on social media — photos or screenshots grabbed wherever inspiration strikes. Second, the output isn't a still image. Doppl converts static images into AI-generated video.

Doppl also brings your looks to life with AI-generated videos — converting static images into dynamic visuals that give you an even better sense for how an outfit might feel.Montana Labs

The screenshot-as-input decision

In May, Google Shopping announced virtual try-on across billions of clothing items via photo upload. That flow starts from structured commerce data: a catalog item with known geometry, angles, and metadata. Doppl explicitly widens the aperture to arbitrary photos and screenshots.

For a frontend team, this is the harder problem. A catalog image is clean and predictable; a screenshot of a friend at a party is not. The garment may be partially occluded, oddly lit, or captured at an angle. Building a product around uncontrolled input means the interface has to set expectations, which is likely why Google frames the whole thing as an early experiment where fit, appearance, and clothing details may not always be accurate.

Why the jump from image to video changes the interaction

Moving the output from a still preview to an animated video is a deliberate interaction choice, not just a rendering upgrade. Google's stated reasoning is that motion gives a better sense of how an outfit might feel — a claim about perception, not accuracy.

That shifts the user's mental model from 'does this render look right' to 'does this look move right.' It's a higher bar. A static composite can hide a lot; animation exposes drape, fit, and continuity across frames. Choosing video anyway suggests Google is betting the emotional payoff of seeing a look in motion outweighs the added surface area for errors — and the save-and-share features are built to turn those videos into social artifacts you send to friends for opinions.

The implication: try-on is becoming a capture-and-share surface, not a checkout step

The Shopping integration sits at the point of purchase. Doppl deliberately detaches try-on from that moment. By accepting photos from anywhere and producing shareable video, it reframes virtual try-on as a standalone consumer app for exploring style rather than a conversion aid bolted onto a product page.

For teams building visual AI features, the takeaway is that the same underlying capability lands very differently depending on where you put the input and output boundaries. Google shipped the constrained, commercial version inside Shopping and the open-ended, social version as a separate Labs experiment. The model work is shared; the product is defined by which inputs you let in and what shape the result takes when it comes out.