Trellis 3D

Trellis 3D is our independent web service for exploring 3D ideas from images or text. The workflow addresses an early design step: turning a visual reference or written concept into an initial asset that can be inspected and developed further. It is intended for designers and creators working on props, product concepts and visual prototypes.

Users begin with an image or text prompt and choose a generation model. The site brings together options including TRELLIS 2, Hunyuan3D, Seed3D, Tripo and Meshy. An interactive browser preview helps inspect the generated form before export. GLB, OBJ, FBX and USDZ exports let users take the initial model into a downstream workflow.

The current product is available on the web. Its frontend uses Next.js, with Codex assisting development. This submission is from the product team, and the service is independent of the official Microsoft TRELLIS research project. We are not presenting it as open-source software or claiming a public API.

Generation is a starting point, not a guarantee of a production-ready mesh. Geometry and topology may need cleanup, and the result still needs to be checked against the requirements of the intended use. A normal signup provides an initial trial; continued generation is paid.

The project link leads directly to the working website. Feedback on the clarity of the image-to-preview workflow, export choices and how clearly these limitations are explained would be useful. The cover image shows the actual website rather than an invented customer result.