A method for audio rendering by an apparatus comprising at least one audio rendering device, the method comprising: extracting (S10) a plurality of frequency band components from an input audio signal; determining (S15) from the device frequency response and the plurality of extracted frequency band components at least one indicator representative of masking frequencies energy, masking frequencies corresponding to frequency bands that are above a frequency threshold; determining at least one correction factor (S16) from the indicator representative of masking frequencies energy; e) for each frequency band, determining a second acoustic level threshold (S20) by modifying (S17) with the correction factor a predetermined first acoustic level threshold associated with said frequency band, and determining a reduction gain from a comparison (S30) between an acoustic level of the extracted frequency band and the second acoustic level threshold associated to said extracted frequency band, and applying (S40) the reduction gain.