---
title: Real-Time Video Try-On
description: >-
  What real-time video try-on is, how generative video models render clothing on
  a live camera feed, and how it differs from photo-based and AR try-on.
shortDefinition: >-
  Generative AI try-on applied to a live video feed instead of a static photo,
  like a mirror that changes your outfit.
relatedTerms:
  - virtual-try-on
  - garment-transfer
  - diffusion-model
  - ar-try-on
heroImage: /glossary-images/photo-vs-realtime-try-on.webp
video: /glossary-videos/real-time-video-try-on.mp4
videoDuration: 31
keywords:
  - real-time virtual try-on
  - video try-on
  - live try-on
publishedAt: '2026-06-11'
updatedAt: '2026-06-11'
noindex: false
canonicalPath: /glossary/real-time-video-try-on
---

## What is real-time video try-on?

**Real-time video try-on** applies generative AI try-on to a live camera feed: the shopper sees themselves on screen wearing a different garment, and the rendered clothing follows their movements like a mirror. It combines the fabric realism of [garment transfer](/glossary/garment-transfer.md) with the interactivity that was previously exclusive to [AR try-on](/glossary/ar-try-on.md).

It is the emerging frontier of [virtual try-on](/glossary/virtual-try-on.md). Photo-based try-on is the production standard today; real-time video generation became credible in 2025-2026 with specialized models (such as Decart's Lucy series) that generate video frames fast enough to feel live.

![A static photo try-on result compared with a real-time video try-on feed](/blog-images/photo-vs-realtime-try-on.webp)

## How it differs from existing methods

| | Photo try-on | AR try-on | Real-time video try-on |
| --- | --- | --- | --- |
| Output | Static generated image | Live 3D overlay | Live generated video |
| Fabric realism | High | Limited | High |
| Interactivity | None (one result) | Full | Full |
| Compute cost | Seconds of GPU per image | Runs on device | Very high (GPU per frame) |
| Production maturity | Standard | Standard for rigid items | Emerging |

## The hard parts

- **Latency**: each frame must be generated in tens of milliseconds; a [diffusion model](/glossary/diffusion-model.md) normally takes seconds per image.
- **Temporal consistency**: the garment must stay identical across frames, with no flickering folds or shifting patterns.
- **Cost**: per-frame GPU generation is orders of magnitude more expensive than one photo render, which currently limits deployment at e-commerce scale.

<Faq>
  <FaqItem question="Is real-time video try-on available for online stores today?">
    Mostly not yet at production scale. The models exist and demos are public, but per-frame generation costs make photo-based try-on the practical choice for storefronts today. Costs follow the usual trajectory of generative AI: down.
  </FaqItem>
</Faq>