Skip to content
← Models
Open weightsStability AI/10 trustMetrics-checked

TripoSR

Single-image to textured 3D mesh in under half a second, MIT licensed.

Visit official page →

About

TripoSR is a fast 3D reconstruction model released in March 2024 (arXiv 2403.02151), built by Stability AI together with Tripo AI (VAST-AI-Research). It uses an LRM-style transformer architecture, an image encoder, an image-to-triplane decoder, and FlexiCubes mesh extraction, totaling around 1B parameters, to turn a single RGB image into a textured 3D mesh in under 0.5 seconds on an RTX 4090.

It is deliberately lightweight: it runs on consumer GPUs with 6GB or more VRAM, includes built-in background removal, and is released under the MIT license, making it one of the most permissively licensed 3D reconstruction models available.

The honest caveats: it produces vertex colors rather than PBR materials, it is image-only with no text-to-3D mode, and geometry for parts of the object the camera never saw is a plausible hallucination rather than a measurement. For higher fidelity, chain it after an image model or step up to Hunyuan3D or TRELLIS.

Specs

Context window

Pricing

Free quota

API access

Weights

Open weights

Released

March 4, 2024

Benchmarks

Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.

Use cases

3d-reconstructionrapid-prototypinggame-assetsar-vr

Trust score — 7/10

/10 trustMetrics-checked

Scored 7/10 (metrics tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 87; Trending score 74/100 from researcher assessment.

Sources