Background: Optically stimulated luminescence (OSL) dosimeters produce a signal linear to the dose, which fades with time due to the spontaneous recombination of energetically unstable electron/hole traps. When used for radiotherapy (RT) applications, fading affects the signal-to-dose conversion and causes an error in the final dose measurement. Moreover, the signal fading depends to some extent on treatment-specific irradiation conditions such as irradiation times. Purpose: In this work, a dose calibration function for a novel OSL film dosimeter was derived accounting for signal fading. The proposed calibration allows to perform dosimetry evaluation for different RT treatment regimes. Methods: A novel BaFBr:Eu 2+ -based OSL film (Z eff, 6 MV = 4.7) was irradiated on a TrueBeam STx using a 6 MV beam with setup: 0 • gantry angle, 90 cm SSD, 10 cm depth, 10 × 10 cm 2 field. A total of 86 measurements were acquired for dose-rates ( Ḋ) of 600, 300, and 200 MU/min for irradiation times (t ir ) of 0.2, 1, 2, 4.5, 12, and 23 min and various readout times (t scan ) between 4 and 1440 min from the start of the exposure (beam-on time). The OSL signal, S( Ḋ, t ir , t scan ), was modeled via robust nonlinear regression,and two different power-law fading models were tested, respectively, independent (linear model) and dependent on the specific t ir (delivery-dependent model). Results: After 1 day from the exposure, the error on the dose measurement can be as high as 48% if a fading correction is not considered. The fading contribution was characterized by two accurate models with adjusted-R 2 of 0.99. The difference between the two models is <4.75% for all t ir and t scan . For different beam-on times, 3, 10.5, and 20 min, the optimum t scan was calculated in order to achieve a signal-to-dose conversion with a model-related error <1%.In the case of a 3 min irradiation, this condition is already met when the OSLfilm is scanned immediately after the end of the irradiation. For an irradiation of 10.5 and 20 min, the minimum scanning time to achieve this model-related error increases, respectively, to 30 and 90 min. Under these conditions, the linear model can be used for the signal-to-dose conversion as an approximation of the delivery-dependent model.The signal-to-dose function,D(M i,j ,t scan ),has a residual mean error of 0.016, which gives a residual dose uncertainty of 0.5 mGy in the region of steep signal fading (i.e.,t scan = 4 min).The function of two variables is representable as a dose surface depending on the signal (M i,j ) measured for each i,j-pixel and the time of scan (t scan ). Conclusions: The calibration of a novel OSL-film usable for dosimetry in different RT treatments was corrected for its signal fading with two different models. A