TRELLIS is Microsoft’s open-source structured 3D latents model that generates high-quality 3D assets from text or images, freely available for research and production.
TRELLIS (Structured 3D Latents for Scalable and Versatile 3D Generation) generates detailed 3D shapes with textures from a single image or text prompt. Its structured latent representation produces cleaner geometry than many alternatives, and the model and code are released openly.
The project supports multiple output formats including radiance fields and meshes, runs on consumer GPUs with reasonable memory needs, and has inspired a wave of open generative 3D tools. For teams that need free, inspectable 3D generation rather than a hosted service, TRELLIS is the leading option. Structured latents produce geometry that needs less repair than many generative approaches, which matters for downstream use. Community optimizations keep improving speed and memory efficiency for local deployment.
TRELLIS is completely free and open source. Costs are limited to the GPU hardware you run it on, and community guides cover efficient local deployment. Being free, total cost is just the GPU you own or rent, making it economical for large batch generation.
TRELLIS is for researchers, developers and 3D teams that want generative 3D without vendor lock-in or usage fees, and for building custom pipelines on an inspectable foundation. It also suits educators teaching generative 3D, since students can experiment freely without accounts or usage caps.
Yes, TRELLIS is entirely free and open source, including for commercial projects under its license terms. You only pay for compute.
TRELLIS is a model and codebase rather than a hosted API, so teams wrap it in their own services. Community projects have built APIs and UIs around it.
TRELLIS trades convenience for freedom: hosted generators are easier, but TRELLIS costs nothing at scale and can be fine-tuned and extended. For open pipelines and research it is the strongest foundation available. TRELLIS is the open foundation to build on; teams wanting managed convenience will prefer hosted services despite the per-generation fees. You will spend more time on setup and engineering than with hosted services, but the payoff is unlimited generation with no per-asset fees and complete control over the pipeline.