TRELLIS

TRELLIS is a local AI model for artists and researchers generating 3D assets from images or text. It supports several 3D representations and can edit generated assets as well as create them.

• Image and text prompts: Generate an asset from a reference image or a written description; the project recommends image-conditioned models for better results.
• Output formats: Produce meshes, radiance fields or 3D Gaussians from a shared representation that captures geometry and appearance.
• Asset editing: Create variations of an object or modify a selected region of a generated asset.
• Training and fine-tuning: Researchers can use the training framework and a curated asset dataset drawn from Objaverse(XL), ABO, 3D-FUTURE, HSSD and Toys4k.

The Python code runs on your own hardware. Linux is the tested platform; Windows setup isn't fully tested. An NVIDIA GPU with at least 16GB of VRAM is required.

Pretrained image-to-3D and text-to-3D models are available through Hugging Face and can load from local files. A Gradio demo provides a browser interface to the locally running model. The repository is open source under the MIT license.