Polycrystalline materials consist an ensemble of grains, and the distribution of their lattice orientations is described as the crystallographic texture. Texture is one of the main factors governing mechanical and functional properties. The evolution of texture is highly sensitive to processing parameters, such as deformation rate, thermal history, strain path etc., making its accurate prediction central to microstructure-sensitive process design. Conventionally, texture is represented through the orientation distribution function (ODF), a probability density over the rotation group SO(3) which is a non-Euclidean manifold. We introduce a generative modeling framework based on normalizing flows that leverages bidirectional transformations to map simple learns latent distributions to complex target distributions, making them well suited for capturing the crystallographic texture. This approach enables both unconditional sampling of complex textures and conditional modeling , enabling both process-structure linkage and inverse design through optimization. The method is able to capture the multimodal distributions and offers mapping between processing conditions and resulting texture. This framework underscores the potential of generative machine learning to advance materials design and microstructural analysis in non-Euclidean domains.