Moonbase59
commented
3 months ago Hmm. A possible reason I could see is that cue_file found a reason to re-analyze the file, then overriding with freshly calculated data. News typically vary in length, maybe that’s it, or—if you pretagged—some diverging parameter between the tags and the station settings (mostly caused by blankskip).
It would be helpful to also see the lines above Metadata added/corrected…, like such:
2024/06/28 22:00:15 [autocue.cue_file:3] Now autocueing: "/var/azuracast/stations/niteradio/media/time/time.niteradio-test-loveit.mp3"
2024/06/28 22:00:15 [autocue.cue_file:3] Blank (silence) skipping active: true, set to 5. s
2024/06/28 22:00:15 [autocue.cue_file:3] Clipping prevention active: true
2024/06/28 22:00:15 [autocue.cue_file:3] Writing tags: false, including ReplayGain: false
2024/06/28 22:00:15 [autocue.cue_file:3] Writing metadata to /tmp/cue_filed1251d.json
2024/06/28 22:00:15 [autocue.cue_file:3] cue_file result for "/var/azuracast/stations/niteradio/media/time/time.niteradio-test-loveit.mp3": {"duration": 7.199999999999999, "liq_cue_duration": 4.9, "liq_cue_in": 2.0, "liq_cue_out": 6.9, "liq_cross_start_next": 6.8, "liq_longtail": false, "liq_sustained_ending": false, "liq_loudness": "-13.99 LUFS", "liq_loudness_range": "5.15 LU", "liq_amplify": "-4.01 dB", "liq_amplify_adjustment": "0.00 dB", "liq_reference_loudness": "-18.00 LUFS", "liq_blankskip": 5.0, "liq_blank_skipped": false, "liq_true_peak": 0.864, "liq_true_peak_db": "-1.27 dBFS"}
2024/06/28 22:00:15 [autocue.cue_file:3] Inserted replaygain_track_gain -4.01 dB and replaygain_reference_loudness -18.00 LUFS
2024/06/28 22:00:15 [autocue.cue_file:3] Given fade-out (0.1 s) < overlay duration (0.1 s), moving cue-out point from 6.9 s to 6.9 s.
2024/06/28 22:00:15 [autocue.cue_file:3] Metadata added/corrected for "/var/azuracast/stations/niteradio/media/time/time.niteradio-test-loveit.mp3":
...plus the lines from _showmeta for this track, to see if the Liquidsoap integration layer did something odd.
Also, if one of the mandatory tags is missing, cue_file might initiate a new analysis. You can check these using cue_file --help:
The absolute minimum set to (possibly) avoid a re-analysis is:
duration, liq_cross_start_next, liq_cue_in, liq_cue_out,
replaygain_track_gain.(These might not always be enough.)
I still think the values from the VCE or advanced tab should override even a new analysis, but to see that, we’d need the output from show_meta (=the final values having been used).
P.S.: Did you ever try to just set liq_fade_in and liq_fade_out on a "news" or "time" file? That would avoid all the extra length calculation stuff. I use that with my jingles and "time" announcements (which of course also vary in length), and it works nicely.
This is my setting for above shown variable-length time.niteradio-test-loveit.mp3:
LordHelmchen666
Not an issue, more a question to understand auto_cue better. I download news file, get the duration with ffprobe and use Azuracast API to set cross_start_next and cue_out with full duration.
But in LS log it seems it don't use this values. Why?