Transcription in human mitochondria is driven by a core apparatus consisting of a Pol A family RNA polymerase (mtRNAP), the initiation factors TFAM and TFB2M, and the elongation factor TEFM. While earlier structures of initiation and elongation complexes provided valuable snapshots, they represent isolated stages of a highly dynamic and multistep process. Critical aspects of mitochondrial transcription - such as DNA recognition and melting, promoter escape, and the release of initiation factors - remain poorly understood. Here, we present a series of cryo-EM structures that capture the transcription complex as it transitions from the initial open promoter complex to the processive elongation complex through intermediate stages. Our data reveal new determinants of promoter specificity, the sequential disengagement of mtRNAP from TFAM and the promoter, the release of TFB2M, and the recruitment of TEFM. Together, these findings provide a detailed molecular mechanism underlying transcription in human mitochondria.