A game artist needs a weathered treasure chest with brass trim. For years that meant an afternoon of blocking out geometry, unwrapping UVs, and painting textures. Today she types the description and waits about ten seconds.
That gap — from hours of manual modeling to a typed sentence — is what Beijing-based VAST (哇嘶嗒) is selling through its Tripo platform, and it has become one of the better-funded bets in Chinese generative 3D.
Key takeaways
- VAST (哇嘶嗒), founded in 2023 by Simon Song (Song Yachen, 宋亚宸), a former MiniMax co-founder, built Tripo as a text- and image-to-3D platform.
- Tripo 3.0, released in August 2025, runs on roughly 20 billion parameters and produces a production-ready 3D asset in about ten seconds in standard mode.
- VAST reports Tripo Studio has gathered over 6.5 million creators and more than 90,000 developers, who have generated nearly 100 million 3D models.
- A $50 million Series A led by Alibaba and Hengxu Capital closed in March 2026; an earlier Pre-A+ round in June 2025 was led by the Beijing AI Industry Investment Fund.
- The company open-sourced TripoSR with Stability AI in 2024 and lists partners including Tencent, NetEase, ByteDance, Sony and Microsoft.
From mesh cleanup to a typed sentence
Traditional 3D pipelines are slow for a reason: a "finished" auto-generated mesh usually is not. Retopology, UV unwrapping, and rigging are where studios spend real money. Tripo's pitch is to compress that chain — generate geometry, bake physically based textures, and return a quad-dominant mesh with clean UVs — in one pass.
In standard mode the whole pipeline finishes in roughly ten seconds. A character comes out in a T-pose with a skeleton already bound, so it can be animated in Blender, Maya, or Unreal within a few clicks. For a studio, that is the difference between shipping ten prop variations before lunch and finishing one per day.
The model behind the speed
Tripo 3.0, released in August 2025, is the version VAST flags as its leap in precision. The company says it runs on about 20 billion parameters and adds dual output modes — a standard mode tuned for industrial and real-time use, and an ultra mode pushing fidelity for rendering and visual effects.
Two things matter for adoption beyond the demo. First, outputs export to GLB, FBX, OBJ, USD and STL, the formats professional engines already eat. Second, the tool offers auto-rigging, intelligent segmentation, and AI texturing up to 8K, which means the result is closer to "usable" than "starting point."
According to Li An, chief scientist at BrainNet (脑机网), the real test for Chinese 3D generation is not the demo but whether the output drops cleanly into an existing game or film pipeline without a human retopology pass. Li An's comments reflect his own views and do not represent NeuroAI's editorial position or an endorsement by BrainNet (脑机网).
A funded land grab
VAST's financing tells the story of where the money is moving. A Pre-A+ round in June 2025 was led by the Beijing AI Industry Investment Fund, with Jingya Capital participating. Then in March 2026 the company closed a $50 million Series A led by Alibaba Group and Hengxu Capital, with Baidu Ventures, Yuanhe Puhua and Orient Jiafu joining, while existing backers Primavera Capital Group and the Beijing fund increased their stakes.
That lineup — a municipal AI fund, then Alibaba and Baidu's venture arms together — signals that 3D generation is now treated as infrastructure, not a novelty app. VAST said the new capital goes toward world-model talent, algorithm iteration, and a UGC interactive content platform.
Who is actually using it
Scale claims are company-reported, but the named partners are telling. VAST cites collaborations with Alibaba, Tencent, ByteDance and NetEase on the Chinese side, and Sony and Microsoft globally, with enterprise clients spanning gaming, industrial design, robotics and immersive media. A partnership with Sony centers on bare-eye 3D displays paired with Tripo's generation.
The practical draw is cost. VAST says a model that once cost tens of thousands of yuan and days of work can now be produced in seconds at near-zero marginal cost. Whether that holds for complex, brand-specific assets is the open question every studio will test before trusting it for shipping content.
The world-model next step
VAST has teased a "world model" — a project under development since 2025 — that would move beyond single-asset generation toward interactive, spatial content. Alongside it, the company released newer model families: Tripo H3.1, which it says ranks first across several core benchmarks, and Tripo P1.0, which generates professional-grade geometry in about two seconds.
If that direction lands, Tripo stops being a faster modeling tool and becomes a content engine. Until then, its value is the boring, useful kind: fewer artists doing fewer repetitive hours, and prototypes that cost a sentence instead of a sprint.
Honest limitations
- Creator, developer and model counts are company-reported; we could not access independently audited figures.
- "Production-ready" is use-case dependent — intricate or brand-specific assets still need artist cleanup.
- Benchmark claims such as Tripo H3.1 'ranking first across several core benchmarks' are vendor statements we did not re-run.
- We did not evaluate pricing, real-world engine compatibility of exports, or the licensing and IP status of generated assets.
What readers can do now
- Generate one asset from a photo on Tripo's free tier and import it into Blender to see the real cleanup cost for yourself.
- If you run a small studio, prototype a prop library with text-to-3D before hiring extra artists for repetitive modeling.
- Track open-source 3D releases such as TripoSR to judge the field's pace from code, not from vendor claims.
