What is TRELLIS?
TRELLIS represents a groundbreaking approach to the creation of 3D assets, crafted by Microsoft Research and featuring a comprehensive Python codebase. By utilizing both text and image prompts, it is capable of generating a diverse array of 3D assets, such as meshes, 3D Gaussians, and radiance fields. This initiative is supported by pretrained models that are conditioned on both images and text, and it includes essential resources like inference pipelines, tools for creating variations of assets, editing features for targeted modifications, and a dataset toolkit. Researchers and developers can either run it on their own Linux machines equipped with a compatible NVIDIA GPU or take advantage of online demos accessible through platforms like Microsoft Foundry and Hugging Face. To ensure successful installation, users should have Python and CUDA configured, along with at least 16 GB of GPU memory for optimal performance as indicated by the tested configuration. The website for the project specifies that its offerings are intended for academic and research purposes, not for commercial use, and it advises users to review the licenses for individual components and repositories before proceeding. Furthermore, the accompanying documentation provides thorough guidance to facilitate an effective setup process for all users interested in leveraging this innovative technology. Overall, TRELLIS stands as a significant advancement in the field of 3D asset generation.
Pricing
Integrations
Screenshots and Video
TRELLIS workflow diagram for 3D asset encoding, structured latent generation, and decoding to 3D outputs.