Decart’s Lucy 2.5 adds live video effects, but physical AI steals the show
▼ Summary
– Decart released Lucy 2.5, an AI video model that generates real-time streaming visual effects, such as fireballs and explosions, without post-production.
– The model provides finer control for placing AI-generated elements and can erase unwanted objects from live videos without visible artifacts.
– Lucy 2.5 enables real-time virtual try-on for ecommerce by merging product images with live camera footage to show how items would look.
– The model’s improved physics engine enhances Decart’s world model simulations, used for training autonomous robots and self-driving cars.
– Decart secured a $300 million funding round backed by Amazon, Nvidia, and others, and partnered with Comcast and AWS to scale its infrastructure.
Artificial intelligence startup Decart is once again redefining the boundaries of live video generation with Lucy 2.5, a model that now brings real-time streaming visual effects to the forefront. For the first time, creators can embed AI-generated VFX directly into live footage as it streams, eliminating the need for post-production work.
According to Decart, Lucy 2.5 stands alone in its ability to produce cinematic effects on the fly. Media production houses and independent artists can now summon realistic fireballs, make characters exhale smoke, burst into flames, or even inflate and explode in a gory mess, all with a simple text prompt. The promise is Hollywood-level effects, limited only by imagination, without the traditional delays of editing.
But Lucy 2.5 isn’t just a tool for creative expression. Its refined grasp of complex physics and visual detail also pushes Decart closer to its larger goals in physical AI. The same engine that delivers cinematic flair can power more accurate world model simulations, which are critical for robotics training. This dual-purpose design means Lucy 2.5 is as much about advancing autonomous systems as it is about dazzling audiences.
What sets Lucy 2.5 apart?
Since launching Lucy 2 just five months ago, Decart has made significant upgrades. The headline feature is the addition of realistic visual effects, allowing users to insert explosions, fire, tsunamis, or completely transform on-screen characters,like turning them into a melting chocolate bar. Creatives can prompt special effects freely, and thanks to an improved physics engine and better prompt adherence, the results look far more convincing than earlier versions.
Precision has also improved. Decart offers users finer control over where and how AI-generated elements appear and behave within a live video. Erasing unwanted objects or characters is now seamless, leaving no visible artifacts. These capabilities are already available through the Decart API.
Advertising and marketing are obvious early targets. With Lucy 2.5, brands can create personalized promotional clips in real time. Shopping experiences could become far more immersive, too. Instead of imagining how a coat or dress might look, users can connect Lucy to a camera, select a product image, and watch as the engine merges them into a high-definition video showing the item in the real world. The implications for virtual try-on and in-store magic mirrors are substantial.
A major step for world models
Lucy 2.5 marks another milestone in Decart’s rapid ascent. Over the past two years, the startup has emerged as a leading force in AI video generation, competing with names like Runway, Luma AI, Pika Labs, and even Google. In December, Decart partnered with Amazon Web Services to integrate Lucy into Amazon Bedrock and Trainium processors. In March, Comcast and Nvidia joined forces with Decart to build a distributed AI grid using Comcast’s edge architecture.
That infrastructure laid the groundwork for Lucy 2, followed by a $300 million funding round in May. Backed by Amazon’s VC arm, Nvidia, Adobe Ventures, Atreides Management, Sequoia Capital, and others, Decart has positioned itself not just as a video AI heavyweight, but as a key innovator in world models and video compute infrastructure.
World models are closely tied to AI video, generating environments that simulate real-world physics. Decart’s Oasis 3 engine has already made strides in robotics training, helping developers fine-tune systems for autonomous robots and self-driving cars. Lucy 2.5 promises to accelerate that work. Its improved precision and visual effects can enhance simulations in clever ways. Developers could test how an autonomous car handles dense fog when camera views fail, or create extreme scenarios like drones avoiding tornados or robots responding to industrial explosions. By combining Lucy 2.5 with Oasis 3, Decart can generate more unpredictable and realistic training simulations, pushing the boundaries of physical AI.
(Source: The Next Web)