We perform radiative transfer simulations for kilonova in various situations, including the cases of prompt collapse to a black hole from neutron star mergers, high-velocity ejecta possibly accelerated by magnetars, and a black hole-neutron star merger. Our calculations are done employing ejecta profiles predicted by numerical-relativity simulations and a new line list for all the r-process elements. We found that: (i) the optical emission for binary neutron stars promptly collapsing to a black hole would be fainter by ⪆1-2 mag than that found in GW170817, while the infrared emission could be as bright as that in GW170817 if the post-merger ejecta is as massive as ≈0.01 M o˙; (ii) the kilonova would be brighter than that observed in GW170817 for the case that the ejecta is highly accelerated by the electromagnetic energy injection from the remnant, but within a few days it would decline rapidly and the magnitude would become fainter than in GW170817; and (iii) the optical emission from a black hole-neutron star merger ejecta could be as bright as that observed in GW170817 for the case that sufficiently large amount of matter is ejected (⪆0.02 M o˙), while the infrared brightness would be brighter by 1-2 mag at the same time. We show that the difference in the ejecta properties would be imprinted in the differences in the peak brightness and time of peak. This indicates that we may be able to infer the type of the central engine for kilonovae by observation of the peak in the multiple band.