<p>Conditional mixture modeling utilizes a parsimonious representation of location parameters for model-based clustering of non-compact clusters. This paper introduces a double-layer conditional mixture model, designed to handle scenarios where non-compact clusters exhibit intricate shapes composed of multiple curved patterns. The novelty of our proposed methodology lies in that mixtures of Gaussian densities are applied to each component’s conditional distributions and the optimal number of sub-components in each conditional distribution is selected. The method automatically addresses situations where a one-to-one correspondence between components and data groups is impractical. A fitting framework based on the expectation-maximization algorithm is outlined. Illustrative examples as well as simulated and real datasets are presented to demonstrate the advantages of our proposal.</p>