TRELLIS

TRELLIS is a 3D asset generation model described in Microsoft’s official project repository, with image- and text-conditioned models, radiance-field, 3D Gaussian and mesh outputs, plus editing and export examples.

Features described on the official site

• The README distinguishes image- and text-conditioned models and describes multiple 3D representations, asset variants and local editing.
• Examples cover GLB mesh and PLY Gaussian exports and a Gradio demo; these examples were not executed here.
• Training code and TRELLIS-500K dataset material are provided; training, inference and dataset use need separate condition checks.

Suggested starting workflow

• Read the project documentation and distinguish image conditioning, text conditioning and output representations.
• Before installation, review GPU, Linux/CUDA requirements and licenses for models, code, submodules and data.
• For later evaluation, use permitted inputs to inspect geometry, textures and exports before deciding downstream suitability.

These are editorial starting suggestions based on product purpose, not workflows tested by this site.

Before you use it

• Not tested hands-on here; feature descriptions come from official material.
• Pricing, regions, languages, commercial use, watermarks and export conditions have not been verified for a single plan.
• The README requires Linux and an NVIDIA GPU with at least 16GB memory; the Windows setup is not fully tested. These are documented prerequisites, not results from this site. The recommendation to generate images with external text-to-image models first is not an included free step; text models and tuning-free multi-image conditioning have stated quality limits. Models and most code are described as MIT, while diffoctreerast/Flexicubes submodules and dataset licenses require separate review; no cost-free or unrestricted commercial workflow is certified.