Real-time geometric processing. Arrangements, booleans, registration, remeshing, and spatial queries — at interactive speed on million-polygon meshes. Exact and robust on real-world data.
Why trueform
Simple code just works. Zero-copy views in C++, NumPy arrays in Python, vectorized NDArrays in the browser. Same engine underneath.
Exact predicates and canonical topology. Handles non-manifold flaps, inconsistent winding, coplanar faces, and pipeline artifacts.
Interactive speed on million-polygon meshes. Benchmarked against MeshLib, VTK, CGAL, libigl, Coal, FCL, and nanoflann.
Spatial trees for collision, distance, ray casting, and k-NN. Connectivity analysis, boundaries, and connected components.
Multi-mesh arrangements and self-intersection resolution. Exact predicates with canonical topology and commutative correctness.
Decimation and isotropic remeshing. Point cloud alignment with ICP, OBB fitting, and rigid registration.
Performance
Integration
cmakeconannugetauto [result, labels, face_labels] =
tf::make_boolean(mesh0, mesh1, tf::boolean_op::merge);pip install trueform(faces, points), labels, face_labels =
tf.boolean_union(mesh0, mesh1)npm install @polydera/trueformconst { mesh, labels, faceLabels } =
tf.booleanUnion(mesh0, mesh1);Research
The algorithms behind the engine — exact arithmetic, spatial hierarchies, robust topology — published and continuously updated.
If the core geometry loop is the main risk, start here.