Translation initiation in human mitochondria relies upon specialized mitoribosomes and initiation factors, mtIF2 and mtIF3, which have diverged from their bacterial counterparts. Here we report two distinct mitochondrial pre-initiation assembly steps involving those factors. Single-particle cryo-EM revealed that in the first step, interactions between mitochondria-specific protein mS37 and mtIF3 keep the small mitoribosomal subunit in a conformation favorable for a subsequent accommodation of mtIF2 in the second step. Combination with fluorescence cross-correlation spectroscopy analyses suggests that mtIF3 promotes complex assembly without mRNA or initiator tRNA binding, where exclusion is achieved by the N-terminal and C-terminal domains of mtIF3. Finally, the association of large mitoribosomal subunit is required for initiator tRNA and leaderless mRNA recruitment to form a stable initiation complex. These data reveal fundamental aspects of mammalian protein synthesis that are specific to mitochondria.