AI startup Decart has rolled out an incremental yet significant improvement to its flagship, real-time video-generation model Lucy, as it pushes to make further inroads with the content creator community.
The new release, called Lucy 2.6, debuted on November 5, building on the live streaming effects that Decart pioneered in the previous iteration of the model. The launch of Lucy 2.5 this summer helped to establish AI video as a tool that enables more precise engagement with consumers, and the company has responded to users’ demands for greater visual fidelity when editing AI-generated content as it streams.
Aesthetic refinements
In a blog post, Decart said it has made a huge effort to refine Lucy 2.6 so as to make it more faithful to the user’s request, no matter if the inputs are a simple text-based prompt or using a reference image upload. “Whether it’s a product shot, a garment, an outfit or a character, the result follows your request much more closely and holds the reference’s detail inside the stream,” the company wrote.
Lucy 2.6, available on the Decart platform now, is enhanced by Decart’s proprietary DOS infrastructure, and most of the performance and stability upgrades stem from that highly optimized architectural foundation, the company said. Existing users will be able to notice a significant improvement in the sharpness of the model’s output, as well as greater accuracy in terms of the finer details of their content.
For instance, when using image references to add digital objects such as a garment to a live feed, the object will look much more realistic, natural and consistent, as if it was actually part of a real scene, Decart said.
Introducing Lucy 2.6.
The world’s leading live video model, now with a major quality upgrade.
Here’s what’s new 🧵
— Decart (@DecartAI) November 5, 2026
Other quality enhancements include more accurate lip syncing and greater realism for both animated and photorealistic AI-generated characters. As a result, creators will be able to generate avatars that represent their facial expressions and other nuances much more accurately.
Finally, Lucy 2.6 supports more powerful prompt understanding, so users will be able to request more detailed, multistep edits in a single text-based prompt. Someone might request the model to generate or edit the background while simultaneously swapping out a character or object, and the command can be executed using a single reference image, Decart said.
Enhancing livestreams on the fly
Decart detailed a number of use cases for Lucy 2.6, and serving the creator economy is one of its primary goals. Social media influencers, bloggers and podcasters are encouraged by Decart to take advantage of its real-time video editing capabilities to get creative and adapt their content on the fly without breaking the flow of their livestreams.
Lucy 2.6 promises to be a powerful tool because the livestreaming ecosystem is all about audience interaction, so the ability to swap character or location mid-video can dramatically increase engagement with viewers. As Decart explains, the idea is to make video engagement more of a two-way street. With audience request-driven broadcasts, viewers can trigger different live visual effects simply by leaving a comment, which prompts the model to switch things up. Alternatively, creators can remain in full control, feeding reference images into the model to transform themselves into different characters, change what products they’re holding, switch up their outfit or teleport themselves into an entirely different location in an instant.
Decart said it has already integrated these features into its Delulu Stream application for social media creators and livestream sellers. Through Delulu Stream, a creator might use Lucy 2.6 to model various outfits they’re promoting. They could start with one, and once they’re done with their sales pitch they can instantly switch to a new outfit without interrupting the stream and immediately strut their stuff for the camera again.
With Lucy 2.6, creators will be able to generate a lifelike replica of almost any product based on just a single reference image. They’ll be able to make the product appear like magic in their hands, or place a prop onto a digital or physical set. Whatever they conjure up, the model will ensure it appears in exactly the right size, with the perfect proportions and finer details like material lighting and shadow perfectly replicated to trick viewers into thinking they’re seeing the real deal.
It’s all about creating more opportunities for content creators. They may not have physical access to the products they’re trying to pitch, but so long as they have a detailed image, that’s really all they need.
AI video finds its niche
Content creators are a primary market for Lucy 2.6, but they’re not the only target. Decart said the model also has applications for more professional video broadcasts. In one example, it showcased how The Famous Group used Lucy 2.6 to generate live visual effects and dynamic scenery changes in mixed reality jumbotron displays during a recent Colorado Avalanche hockey game.
Retailers like Under Armour are also using Lucy 2.6 to show shoppers how good they might look while wearing its latest clothes and accessories as they stroll past its store.
By putting such powerful editing capabilities at the fingertips of every content creator, Decart once again demonstrates that the future of AI video lies in real-world functionality rather than the flashy gimmicks seen with earlier iterations of the technology.