# Meshy T2: Native 3D Mesh Generation with Flow Matching
Adapted from Rendong Liang’s article on the Meshy blog . When you evaluate a generated 3D asset, the preview image answers only part of the question. What does the wireframe look like? Are separate components still separate? Can the geometry fit your application’s budget? Meshy T2 explores these questions by generating mesh geometry and connectivity directly. Represent the mesh as a set…
Meshy T2 is a novel method for generating 3D mesh geometry and connectivity directly, without the need for separate generation and stitching processes. Rather than generating a wireframe image, Meshy T2 represents the mesh as a set of vertices, edges, and faces using a vertex-set variational autoencoder. This allows for more control over the structure of the generated mesh.
The mesh generation process uses two flow-matching stages. The first, voxel flow, uses a reference image to establish a coarse occupancy scaffold. The second, mesh flow, generates vertex tokens conditioned on the image, scaffold, and a requested vertex budget. These tokens are then turned into geometry and connectivity by the mesh decoder during a single reconstruction step.
An important aspect of Meshy T2 is its ability to generate disconnected components within a single asset. This is achieved without needing to stitch together separate generation results. The method also supports control over the complexity of the generated mesh, with budgets ranging from 500 to 4,000 vertices. Padding tokens are used to discard surplus vertices during decoding.
In terms of generation speed, the published paper reports a median end-to-end image-to-mesh time of 6 seconds. However, an engineering-optimized version is reported to take approximately 1 second, or roughly 2 seconds including the WebApp/API service round trip. These times are based on different execution settings and should not be viewed as universal latency guarantees for every input or deployment.
When evaluating Meshy T2 for a specific use case, it is important to inspect the output beyond just the rendered visual. Factors to consider include whether important shapes survive at the intended budget, whether the generated connectivity matches the object's structure, whether the asset behaves correctly after being imported into target tools, and how much additional editing may be required.
The full Meshy article provides architecture diagrams, wireframe examples, and budget comparisons to assist in this evaluation process.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.