On Sat, Feb 2, 2019 at 3:52 PM Romain Beauxis <[email protected]> wrote: > You're using smart_crossfade, right?
Thank you for asking. I have not experimented with smart_crossfade. Incidentally, has some code changed regarding "amplify"? My playlist contains gains (see below) in dB but, as of 1.4.0-alpha1, those requesting amplification (not attenuation) distort heavily. They do not distort in 1.3.6. Instead of using smart_crossfade, each track is individually transferred from its original file to a mezzanine format and, in doing so, a script detects the position in each track where (1) its opening audio exceeds a certain threshold, and (2) the last point in the track where the EBU R.128 volume dips lower than a set level below the track's average volume (again, EBU R.128 measure). The script also detects the additional gain necessary to achieve similar levels, and annotates this using liq_cue_in, liq_start_next, etc. Code for annotation: https://github.com/Warblefly/TrackBoundaries/blob/master/cue_playlist.py This information then feeds liquidsoap's "crossfade" function, with fade_out and fade_in set to 0.01 (very short), so that the tracks themselves do their own fading, if any exists. In this way, smart_crossfade isn't needed because (1) the incoming and outgoing volume of each track determines the exact moment the next track begins, and (2) the output audio processing allows an incoming loud track to over-ride an outgoing quiet track. However, some tracks fade out over a very long time, and I ought to fade those more quickly e.g. over 5". I realise you have worked hard on the smart_crossfade algorithm, and I need to read up on it more than I have done. http://warblefly.sytes.net:8000/audio.aac with best wishes, JW _______________________________________________ Savonet-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/savonet-users
