Turbulence is not Gaussian. In this talk I will discuss the recently developed framework of Ambit Stochastic and show how it can be used to construct non-Gaussian random vector fields that are easy to simulate from, the main application being the velocity vector field in a turbulent flow. First we see how to reproduce the correlation structure, i.e., the spectral tensor in a homogeneous, isotropic, incompressible vector field. Then we see simple ways of introducing inhomogeneity and anisotropy. Important features of turbulence are the non-Gaussian (peaked, heavy-tailed and slightly skewed) distributions of velocity increments and the intermittency of the energy dissipation. Using the generalized hyperbolic distribution we see how to produce non-Gaussian increments, and using stochastic volatility modulation we see how to produce intermittency.