Type a sentence. Get a rigged, textured, game-ready 3D model before your coffee cools. That is the whole pitch - and roughly 100 million assets say it works.
A close read of a company that took the slowest, most manual corner of digital creation - 3D - and pointed a foundation model straight at it.
Here is a fact that sounds obvious once you say it out loud: almost everything on a screen that isn't a flat photo or a block of text is a 3D model that somebody built by hand. The character in the game. The sneaker you rotate before buying. The chair in the furniture app. Each one is the product of a skilled artist, specialized software, and - this is the important part - a lot of time. 3D is the slow corner of the internet, and it has stayed slow while text and images got automated.
Tripo AI is a bet that this corner is about to get very fast. The company, operated by a startup called VAST, makes software that generates 3D models from a text prompt or a photo. You describe an object, or upload an image, and a few seconds later you have a mesh you can drop into a game engine. The interesting thing about Tripo isn't just that it can do this - several companies can - but that it treats the model as the beginning rather than the end. A raw 3D mesh is a bit like a raw slab of marble: technically the sculpture is in there, but you still have work to do. Tripo tries to do the rest of the work too.
The result is that a task which used to cost an artist a day or a week now costs a few seconds and a few cents. When something gets that much cheaper, you don't just do the old thing faster - you start doing new things that were never worth the old price. That is the part worth watching.
Tripo AI aims to make high-quality spatial creation accessible to everyone - enabling users to create, edit, experience and interact with 3D content and persistent worlds.- Company Mission Statement, 2026
The consumer-facing side is Tripo Studio, a browser workspace. The developer side is a REST API that roughly 90,000 developers use to wire 3D generation into their own products. Both cover the same pipeline, which is the point - it is a full workflow, not a party trick.
Text-to-3D, image-to-3D, and multiview-to-3D, producing models up to 500K polygons. The 3.0 release is the first the company calls genuinely "product ready."
Physically based rendering textures now up to native 8K, plus a "magic brush" for editing surfaces directly instead of re-rolling the whole model.
Automatic skeletons for avatars and creatures, so a static sculpture becomes something that can actually move and animate.
Intelligent Part Segmentation V2 breaks a model into editable pieces at chosen granularity - 3-6, 6-15, or 15+ parts.
Retopology and low-poly optimization for XR and gaming, where every polygon has to earn its place in the frame budget.
One-click export to USD, FBX, OBJ and STL. It fits the toolchain you already have rather than asking you to adopt a new one.
Bonus: six one-click stylization looks, including cartoon, clay, steampunk - and, yes, a Christmas theme.
A useful way to read a creative-AI company is not "how good is the demo" but "how many people showed up." Tripo's adoption is its most defensible asset.
Bars are illustrative relative scale, not exact ratios. Figures are company-reported and press-reported approximations.
Tripo's founder and CEO, Simon Song, has an unusual resume for a 3D-graphics entrepreneur. Born in 1997, he studied Middle East politics at Johns Hopkins before joining SenseTime, where one of the business lines involved 3D animation. He later co-founded MiniMax, one of China's best-known AI unicorns.
Then he left to start something harder. Language models were already crowded; Song's thesis was that the next big modality wasn't text or images but 3D - and eventually whole interactive worlds. VAST was founded in 2023 to chase that idea, with co-founder Guoli Su, and teams that grew across San Francisco, Beijing and Hangzhou.
The framing he uses is deliberately unglamorous: let everyone create 3D content with zero barriers and zero cost. It is a distribution-first idea more than a graphics-first one. Make the thing free and easy enough, and millions of people will do the marketing for you - which is more or less what happened.
By 2025 he was describing a business that had crossed into real revenue, on the order of a million dollars a month, while still being early. For an infrastructure-flavored company selling into games, XR and platforms, that is the point where investors start paying very close attention.
VAST aims to let everyone create 3D content with zero barriers and zero cost.- Simon Song, Founder & CEO
Funding arrived in a rush through 2026, with backers spanning big tech, automotive, gaming and finance - a mix that reads less like a bet on a chatbot and more like positioning for spatial computing.
| Round | Amount | Date | Notable Backers |
|---|---|---|---|
| Funding round | $50M | Mar 2026 | Alibaba, Baidu Ventures |
| Series A+ / A++ | ~$200M | Jun 2026 | Strategic & financial investors |
| Series A3 | $150M+ | Jul 2026 | Geely Capital, Giant Network, 4399, Fosun, CoStone |
| Total (reported) | ~$400M | 2023-2026 | Across all rounds |
Round labels and totals are as reported across public press; some figures may overlap or be partially disclosed.
The obvious next question for any 3D-generation company is: what happens after you can make one good object? Tripo's answer is Project Eden, a research program aimed at persistent, reusable, multiplayer interactive environments. It uses a three-layer architecture that separates world state from visual rendering - the idea being that a world can persist, be edited, and hold many people at once, rather than being a static file you download.
Environments that keep their state over time instead of resetting - a prerequisite for anything that feels like a place rather than a scene.
Separating world state from rendering means the environment can be edited without re-generating everything from scratch.
Concurrent multiplayer interaction - the difference between a generated diorama and a shared, living space.
The long-term vision Song has floated is almost cheeky: a "TikTok for 3D," where user-generated interactive worlds spread the way short videos do now. Whether that consumer future arrives is an open question. The infrastructure being built toward it is not hypothetical.
Former MiniMax co-founder Simon Song starts VAST to build AI-native 3D generation.
Tripo Studio and the developer API bring text- and image-to-3D generation to creators and developers.
Users pass multiple millions, revenue reaches roughly $1M/month, and open releases like TripoSG expand the ecosystem.
Alibaba and Baidu Ventures back a $50M round as Tripo H3.1 and P1.0 ship with native 8K textures.
Back-to-back rounds fund a push from single 3D assets toward interactive world models.
Millions of individual creators use the free and paid web app. On the enterprise side, reported API partners include Tencent, ByteDance, HTC and Stability AI - the kind of names that don't integrate a startup's API for fun.
Tripo sits among 3D-generation rivals like Meshy, Luma AI, Rodin/Hyper3D, Kaedim and Spline, plus in-house efforts from Tencent (Hunyuan3D), Google and Nvidia. The moat is adoption and the full pipeline, not any single model.
Text and image to 3D, rigging, texturing and export walkthroughs.
The founder on world models, and the "TikTok for 3D" vision.
Auto-rigging, stylization and retopology, step by step.
Docs and reference for the Tripo generation API.
It generates production-ready 3D models from text prompts, images or multiview inputs, and handles texturing, rigging, retopology, segmentation and export - available as a web app (Tripo Studio) and a developer API.
It was founded in 2023 by Simon Song, a former co-founder of AI unicorn MiniMax, under the company VAST, with co-founder Guoli Su.
Close to $400 million total, including $50M in March 2026 (Alibaba, Baidu Ventures), roughly $200M in June 2026 (Series A+/A++), and $150M+ in a July 2026 Series A3.
More than 6.5 million creators and around 90,000 developers, with reported enterprise partners including Tencent, ByteDance, HTC and Stability AI.
Other 3D-generation tools such as Meshy, Luma AI, Rodin/Hyper3D, Kaedim and Spline, plus 3D efforts from Tencent, Google and Nvidia.
Share this profile
Sources: company site, GlobeNewswire, PR Newswire, Unite.AI, SiliconANGLE, SCMP, Forbes, VoxelMatters. Figures are approximate and press-reported.