Quick answer: What is Decart Lucy 2.5, and does it actually work?

Lucy 2.5 is a real-time AI video transformation model developed by Decart. It is designed to modify live video, including backgrounds, characters, objects, and visual styles. I tested approximately 10 to 15 effects in its free public demo in October 2026. While the first few worked reasonably well, the video quality was poor, the effects became increasingly glitchy, and some transformations failed to track my face correctly. The technology sounds promising, but based on this single testing session, the demo didn’t make me want to explore the platform any further.

Why I wanted to try Lucy

Amazon is reportedly in talks to acquire AI startup Decart for around $7 billion, according to Bloomberg (https://finance.yahoo.com/technology/ai/articles/amazon-talks-buy-ai-startup-204652661.html), citing The Wall Street Journal. Nothing is finalized yet, but that’s quite a number. Naturally, I got curious.

I’ve been testing AI tools for a project called Tried, Actually, partly because I want to see whether any of them could eventually help me create better videos. Mostly, though, I’m just curious about where all this technology is going. That’s why I started the project in the first place: I try things myself and share what works, what doesn’t, and what’s worth paying attention to.

After recently testing ChatGPT’s virtual try-on feature (https://triedactually.com/en), where some results looked nothing like the original person, I was particularly curious to see how AI would handle live video. So when I read about Lucy and its ability to transform live video in real time, I wanted to see how far the technology had come.

And when a company potentially worth billions offers a free demo, of course I’m going to try it.

What is Decart, and what does Lucy actually do?

Decart is an AI startup founded in 2023. Its work goes well beyond funny video effects. The company develops technology to make AI systems faster and more efficient, alongside models that can generate and modify visual environments in real time.

One of its better-known projects is Oasis, an AI-generated interactive world where the environment responds to what you’re doing. Another is Lucy, a real-time video transformation model. The idea is genuinely interesting: changing your surroundings during a livestream, virtually trying on clothes, adding objects to a scene, or creating visual effects without spending hours editing footage afterward.

According to Decart, the technology has potential applications in advertising, gaming, social media, virtual try-on, and more. That’s exactly the kind of thing I’d like to use for my own content someday. Amazon’s reported interest concerns a much broader AI business, but Lucy is one of the company’s most accessible tools for ordinary people to try.

Nobody is suggesting Amazon wants to spend $7 billion on a collection of video filters. Still, I wanted to see what this technology looked like outside a polished promotional video.

My Lucy 2.5 test: from curiosity to frustration

The first thing I saw on the Lucy demo page was a queue. I was number six, with an estimated waiting time displayed on the screen. The estimate changed twice while I waited, but after about two and a half minutes, I was in. Not a big deal. It’s a free demo, and I can wait a couple of minutes.

The video quality, however, was my first disappointment. I looked like I was filming myself on a toaster.

Then I started trying the effects, 10 to 15 of them in one session. The first few worked reasonably well, although they weren’t particularly impressive. But the more I tried, the worse things got. The video started freezing badly, and some effects lost track of my face altogether. At one point, Lucy seemed to invent an entirely different woman in my kitchen and apply the effect to her instead of me. Apparently, my kitchen had more people in it than I knew.

Before opening the demo, I’d read about Lucy’s ability to transform environments in real time. In my head, I imagined sitting in my kitchen and suddenly appearing on a beach. Then perhaps in a forest. Ideally, with the kind of realism that makes you look twice. Something I could eventually imagine using in my own videos.

What I got instead was a strange little green alien baby, a mysterious woman in a bathtub with rubber ducks, a blonde wig, and various other effects that reminded me of Snapchat around 2016. And yes, Snapchat already had face filters back then.

The biggest problem wasn’t that the effects looked silly. I don’t mind silly. Some of the best social content is silly. The problem was that the demo gave me almost no sense of what Lucy could actually become. I kept trying more effects, hoping the next one would show me something impressive. Instead, it only got glitchier, and by the end, I was more frustrated than curious.

Watch my Lucy 2.5 test on YouTube Shorts

Isn’t a demo supposed to make you want the product?

A free public demo isn’t necessarily representative of a finished commercial product. There may be limits on processing power, available features, or how many people can use it at once. I don’t know what caused the problems during my session, and I won’t pretend I do. I only tested Lucy once, so I can’t say whether everyone would have the same experience.

Still, a demo should make you think, “Wow, I want this.” Mine was underwhelming from the start and got worse toward the end. I closed the page wondering what exactly I’d just tested, and with no desire to explore the platform any further.

Here’s the frustrating part: what Lucy is designed to do still sounds interesting. Decart describes changing backgrounds, swapping characters, adding or removing objects, and applying visual effects to live video. If that works smoothly and looks convincing, I can see real uses for creators: swapping a background without a green screen, or running a product demo that reacts instantly to what someone does on camera.

Lucy 2.5 is already available to developers through Decart’s platform. What I’d really like to see is a more polished, accessible experience for regular users. I’d gladly test one, though I couldn’t find a confirmed release date for such a version.

My verdict: I expected more

I wanted to like Lucy. A real-time AI tool that can transform entire scenes while you’re filming sounds exactly like the kind of technology worth keeping an eye on. Unfortunately, what I got was a glitchy, low-quality stream with effects that felt years behind the pitch.

Maybe I caught the demo on a bad day, or maybe the commercial version is dramatically better. Either way, with Amazon reportedly considering a multibillion-dollar acquisition, the gap between the technology’s ambitions and this particular demo is hard to miss.

I’ll happily give Lucy another chance when there’s something more convincing to test. For now, though, I’ll stick to my real kitchen.

At least I know who’s in it.

Sources and further reading

Bloomberg via Yahoo Finance: Amazon Is in Talks to Buy AI Startup Decart — https://finance.yahoo.com/technology/ai/articles/amazon-talks-buy-ai-startup-204652661.html

Decart: Official website — https://decart.ai/

Decart: Lucy — https://decart.ai/lucy

Lucy 2.5: Public demo — https://lucy.decart.ai/

Decart: Research and technology — https://decart.ai/research

Tested independently using the free public Lucy 2.5 demo in October 2026. This article reflects one testing session and my personal experience.