Mitochondria have their own genome, their own translational machinery and their own mRNAs which code a handful of proteins. Most of the proteins come from the cytoplasm, but some get translated inside the mitochondria. And this is done in amazingly weid way...
Bacteria - mitochondrial ancestors - have 3 initiation factors, IF1, 2 and 3, and all of these are absolutely necessary for the bacterial viability.
Mitochondria do not have IF1 (all mitochondria), and unlike mammals, yeast ones do not have IF3! Also they seem to use mitochondria-specific initiation factor AEP3. Do mammals have AEP3 homologue? Worth checking... To make things more complicated, mitochondria have their own special mRNA-specific factors... and again, there is a catch. Yeast mRNA have long 5' UTRs (untranslated regions), and mammals have short, they basically have leaderless mRNAs (I'd love to see a good reference for that! THIS is a little bit out of date...) - therefore I would expect that these mRNA-specific IFs work differently in yeast in mammals. So yeast and mammalian mitochondrial translation seems to have very, very different translational apparatus. Isn't is weird?
Another amazing thing is that translation and protein localization are tightly linked in mitochondria. mRNA-specific initiation factors (i.e. Pet111p) and above-mentioned AEP3 interact with the mitochondrial membrane (most of the mitochondrially-translated proteins are membrane proteins), so that translation is localized where the proteins should go.