World Labs, the startup based by AI pioneer Fei-Fei Li, is launching its first industrial world mannequin product. Marble is now accessible by way of freemium and paid tiers that allow customers flip textual content prompts, photographs, movies, 3D layouts or panoramas into editable, downloadable 3D environments.
The launch of the generative world mannequin, first launched in restricted beta preview two months in the past, comes slightly over a 12 months after World Labs got here out of stealth with $230 million in funding, and places the startup forward of rivals constructing world fashions. World fashions are AI methods that generate an inside illustration of an atmosphere, and can be utilized to foretell future outcomes and plan actions.
Startups like Decart and Odyssey have launched free demos, and Google’s Genie continues to be in restricted analysis preview. Marble differs from these — and even World Labs’s personal real-time mannequin, RTFM — as a result of it creates persistent, downloadable 3D environments fairly than producing worlds on-the-fly as you discover. This, the corporate says, ends in much less morphing or inconsistency, and lets customers export worlds as Gaussian splats, meshes or movies.
Marble can be the primary mannequin of its variety to supply AI-native modifying instruments and a hybrid 3D editor that lets customers block out spatial buildings earlier than AI fills within the visible particulars.
“This is a brand new category of model that’s generating 3D worlds, and this is something that’s going to get better over time. It’s something we’ve already improved quite a lot,” Justin Johnson, co-founder of World Labs, informed TechCrunch.
Final December, World Labs confirmed how its early fashions might generate interactive 3D scenes primarily based on a single picture. Whereas spectacular, the considerably cartoonish scenes weren’t absolutely explorable since actions had been restricted to a small space, and there have been occasional rendering errors.
In my trial of the beta preview, I discovered Marble generated spectacular worlds from picture prompts alone — from game-like environments to photorealistic variations of my lounge. Scenes morphed on the edges, although that’s apparently been improved in at this time’s launch. That stated, a world I’d generated within the beta utilizing a single immediate appeared higher and matched my intent extra carefully than the identical immediate does now.
Techcrunch occasion
San Francisco
|
October 13-15, 2026
I haven’t but examined the modifying options, although Johnson says they make Marble sensible for near-term gaming, VFX and digital actuality (VR) tasks.
“One of our main themes for Marble going forward is creative control,” Johnson stated. “There should always be a quick pathway to generate something, but you should be able to dive even deeper and get a lot of control over the things that you’re generating. You don’t want the machine to just take the wheel and pull all that creativity away from you.”

Marble’s tackle inventive management begins with enter flexibility. The beta solely accepted single pictures, forcing the mannequin to invent unseen particulars for a 360-degree view. With the complete launch, customers can now add a number of pictures or quick clips to indicate an area from totally different angles and have the mannequin generate pretty reasonable digital twins.
Then now we have Chisel, an experimental 3D editor that lets customers block out coarse spatial layouts (suppose partitions, packing containers, or planes) after which add textual content prompts to information the visible fashion. Marble generates the world, decoupling construction from fashion — much like how HTML supplies the construction of a web site and CSS provides in shade. Not like text-based modifying, Chisel allows you to instantly manipulate objects.

“I can just go in there and grab the 3D block that represents the couch and move it somewhere else,” Johnson stated.
One other new function that provides you extra modifying management is the flexibility to increase a world.
“Once you generate a world, you can expand it up to once,” Johnson stated. “When you move to a piece of the world that’s starting to break apart, you can basically tell the model to expand there or generate more world in the vicinity of where you currently are, and then it can add more detail in that region.”
Customers who wish to create extraordinarily giant areas can mix a number of worlds with “composer mode.” Johnson demonstrated this for me with two worlds he had already constructed – a room fabricated from cheese with grape chairs, and one other of a futuristic assembly room in area.
The trail to spatial intelligence

Marble is accessible by way of 4 subscription tiers: Free (4 generations from textual content, picture, or panorama), Customary ($20/month, 12 generations plus multi-image/video enter and superior modifying), Professional ($35/month, 25 generations with scene enlargement and industrial rights), and Max ($95/month, all options and 75 generations).
Johnson thinks the preliminary use instances for Marble shall be gaming, visible results for movie, and digital actuality.
Sport builders have combined emotions in regards to the tech. A current Sport Builders Convention survey discovered a 3rd of respondents believed generative AI has a adverse influence on the video games business – 12% greater than the survey indicated 12 months earlier. Mental property theft, vitality consumption and a lower in high quality from AI-generated content material had been among the many prime issues aired. And final 12 months, a Wired investigation discovered recreation studios like Activision Blizzard are utilizing AI to chop corners and fight attrition.
In gaming, Johnson sees builders utilizing Marble to generate background environments and ambient areas after which importing these property into recreation engines like Unity or Unreal Engine so as to add interactive components, logic and code.
“It’s not designed to replace the entire existing pipeline for gaming, but to just give you assets that you can drop into that pipeline,” he stated.
For VFX work, Marble sidesteps the inconsistency and poor digital camera management that plague AI video mills, per Johnson. Its 3D property let artists stage scenes and management digital camera actions with frame-perfect precision, he stated.
Whereas Johnson stated World Labs isn’t specializing in digital actuality (VR) purposes proper now, he famous the business is “starved for content” and excited in regards to the launch. Marble is already appropriate with the Imaginative and prescient Professional and Quest 3 VR headsets, and each generated world will be seen in VR at this time.
Marble might also have potential use instances for robotics. Johnson famous that in contrast to picture and video era, robotics doesn’t get pleasure from a big repository of coaching information. However with mills like Marble, it turns into simpler to simulate coaching environments.
Based on a current manifesto by Fei-Fei Li, CEO and co-founder of World Labs, Marble represents step one in the direction of creating “a truly spatially intelligent world model.”
Li believes “the next generation of world models will enable machines to achieve spatial intelligence on an entirely new level.” If giant language fashions can train machines to learn and write, Li hopes methods like Marble can train them to see and construct. She says the flexibility to know how issues exist and work together in three-dimensional areas can ultimately assist machines make breakthroughs past gaming and robotics, and even into science and medication.
“Our dreams of truly intelligent machines will not be complete without spatial intelligence,” Li wrote.
Obtained a delicate tip or confidential paperwork? We’re reporting on the interior workings of the AI business — from the businesses shaping its future to the individuals impacted by their selections. Attain out to Rebecca Bellan at [email protected] or Russell Brandom at [email protected]. For safe communication, you’ll be able to contact them by way of Sign at @rebeccabellan.491 and russellbrandom.49.
