Meshy 4 is a generative 3D foundation model developed by Meshy, designed to produce high-fidelity 3D assets from text prompts and images. Released in August 2024, the model focuses on significant improvements in geometric precision, particularly for hard-surface modeling. It addresses common issues found in earlier generative 3D versions, such as surface irregularities, by producing cleaner lines, sharper corners, and flatter surfaces suitable for professional workflows in gaming, film, and architectural visualization.
The model introduces a redesigned Text-to-3D workflow that separates the generation process into two distinct stages: Modeling and Texturing. In the modeling stage, the system generates four high-quality untextured meshes based on a text prompt, providing a structural foundation for the user to review. Once a mesh is selected, the texturing stage applies high-resolution materials. This modular approach is intended to prevent the final output from deviating significantly from the initial preview, a common challenge in previous "coarse-to-fine" iterations.
Key capabilities include enhanced Image-to-3D conversion, which allows for the creation of 3D objects from a single reference image with improved structural inference. The platform allows users to switch between the Meshy-4 algorithm for maximum detail and the Meshy-3 Turbo algorithm for faster, organic modeling. To improve the creative workflow, the model supports a Retry function that allows for rapid regeneration of assets without extra overhead if the initial preview does not meet expectations.