In this paper we use a large data set of seismometric and accelerometric recordings (3168 vertical and 1402 for each of the horizontal components) collected by various networks in the eastern Alps to estimate empirical ground-motion attenuation relations valid in the magnitude range 2.5–6.3 for distances of up to 130 km. Relations are developed for horizontal and vertical peak ground acceleration and velocity, Arias intensity, response spectral acceleration, and Fourier amplitude for 46 periods between 0.1 and 2 sec. We adopt a simple magnitude-dependent attenuation model predicting closer attenuation curves and decreasing distance attenuation for increasing magnitude. We use truncated regression analysis to deal with the problem of non-triggering stations; with respect to other approaches, our solution does not require any discarding of data or the knowledge of which stations did not trigger. Our relations agree with others suitable for the study region for magnitudes greater than 5.8, while they have a faster attenuation for lower magnitudes. In an attempt to reduce the standard deviation, we have introduced soil classification into the relation, as well as a characterization of sites by means of H/V ratios. These ratios were estimated either by earthquakes (receiver function) or by the background seismic noise (Nakamura’s ratios). Similar to findings in other studies, the improvement obtained with soil classification is negligible, while H/V ratios reduce the standard deviation by up to 22%. Finally, we have investigated the implications of these new relations for seismic hazard assessment considering two sites of interest located in northeastern Italy (Udine and Gemona). In both cases, the difference in the hazard estimates for return periods shorter than 100 years is notable with respect to all other available relations.