Seismic risk management of building portfolios requires a reliable evaluation of earthquake-induced losses. This is commonly performed using consequence models linking structure-specific damage states (DSs) experienced by a building to a given loss metric (or decision variable). This study demonstrates a simulation-based procedure that derives refined probabilistic consequence models considering two essential loss metrics: direct-loss and repair-time ratios (repair cost or time normalised by the corresponding reconstruction values). Nine case-study reinforced concrete frames with various heights and design-code levels are developed to represent common residential buildings in Italy and the Mediterranean region. The proposed procedure starts by defining building-level, structure-specific DSs that reflect the increasing structural and nonstructural damage for the nine frames. Their seismic response is then assessed by analysing two-dimensional nonlinear numerical models and deriving building-level fragility relationships. Next, component-based direct-loss and repair-time analysis is conducted via the FEMA P-58 methodology, which computes such metrics at multiple ground-shaking intensities using Monte Carlo sampling. The consequence models are finally characterised by fitting probabilistic distributions to the direct-loss and repair-time realisations after conditioning them on the respective global DSs sustained by each case-study frame. This procedure enables deriving enhanced consequence models that can be easily implemented in risk analysis of building portfolios to obtain quick loss estimates. This study finally sheds some light on the possibility of correlating repair time to direct loss, which might be useful in estimating indirect losses resulting from downtime, particularly in cases where repair-time data or models are unavailable.