If you saw "aa1.hair.v1" in:
Unlike volumetric approaches that operate on voxel grids, AA1 processes hair geometry as a sequence of points. We employ a recurrent neural network (RNN) structure combined with a graph convolutional network (GCN) to embed the local and global features of hair strands. This preserves the topological relationship between neighboring strands, which is crucial for avoiding the "stringy" artifact. aa1.hair.v1
Historically, hair modeling relied on procedural noise functions or manual groom tools. While precise, these methods do not scale well for large datasets or real-time generation. If you saw "aa1