A method to determine the kinetic freeze-out temperature in heavy-ion collisions from measured yields of short-lived resonances is presented. The resonance production is treated in the framework of thermal model with an evolution between chemical and kinetic freeze-outs. The yields of many short-lived resonances are suppressed at $T = T_{\rm kin} < T_{\rm ch}$. A fit of $T_{\rm kin}$ allows to describe the abundances of both, the stable hadrons and the short-lived resonances, like $\rho^0$ and $K^{*0}$, as measured by the ALICE collaboration at the LHC as a function of centrality. This allows to extract $T_{\rm kin}$ from the measured hadron and resonance yields alone, independent of assumptions about the flow velocity profile and the freeze-out hypersurface. The extracted $T_{\rm ch}$ values exhibit a moderate multiplicity dependence whereas $T_{\rm kin}$ drops, from $T_{\rm kin} ≃ T_{\rm ch} ≃ 155$ MeV in peripheral collisions to $T_{\rm kin} ≃ 110$ MeV in 0-20% central collisions. Predictions for other short-lived resonances are presented. A potential (non-)observation of a suppressed $f_0(980)$ meson yield will allow to constrain the lifetime of that meson.