The purpose of this paper is to develop an explicit construction of consistent utilities, using the stochastic flows approach developed in El Karoui & Mrad (2013, 2020). Starting from a family of utility functions indexed by some parameter α (for example, the risk aversion coefficient or any other parameter), the idea is to randomize α and to construct nonstandard stochastic utility processes. Two approaches are developed. The first one consists of building directly from the class {Uα,α∈ℝ} a global utility U as a sup-convolution. The second approach which is very different, consists to define from the class (Xα,Yα)α∈ℝ of monotonic optimal processes, associated with the class {Uα,α∈ℝ}, a global pair (X∗,Y∗) as a mixture. The nonstandard stochastic utility is then obtained by composing stochastic flows and is interpreted as the aggregate utility of all considered agents.