Mitochondria serve as energy-generating factories and as modulators for cellular homeostasis. Natural evolution relocated most of the mitochondrial genome into the cell nucleus, except for 13 protein-coding genes and 24 non-coding genes that assist protein translation in mammalian cells. Hundreds of potential pathogenic variants of mitochondrial DNA (mtDNA) have been linked to mitochondrial diseases that typically manifest as severe conditions, which can often be fatal. No curative treatments for mitochondrial diseases are currently available, in part because of the shortage of suitable cellular and animal models. Recently, double-stranded-DNA deaminase (DddA)-derived cytosine base editors (DdCBEs) have been developed to catalyse site-specific C:G to T:A conversions in mtDNA1. These base editors allow for the installation of pathogenic mtDNA variants in cells, mice, zebrafish and rats1,2,3,4,5. However, off-target edits detected in the mitochondrial and nuclear genomes in these cellular and animal models raised concerns about the specificity of the DdCBEs2,3,4,5,6,7. In addition, cost and time constraints made the individual installation of hundreds of pathogenic variants in cells and animals a formidable challenge. Instead, knocking out mitochondrial protein-coding genes would be a viable alternative to the modelling of mtDNA dysfunction in vitro and in vivo. Reporting in Nature Biomedical Engineering, Michal Minczuk and colleagues now describe a library of improved mitochondrial base editors that can be used to precisely and efficiently ablate each of the 13 mitochondrial protein-coding genes8.