Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[halluci293]

I'm not really sure what triggers this, but I've encountered several cases where audio analysis on video files (full length movies) returns the maximum values for R128 (47.0 LU) and ReplayGain (52.00 dB). This is hugely problematic because I have my library run analysis automatically on import, and if I'm not careful I blow my speakers/eardrums out when I start playing one of these files (which have functional audio tracks that at least seem normal on playback). If I manually edit the tags to remove volume leveling, playback levels are perfectly fine and matches levels I get when playing back other files that returned reasonable audio analysis levels. Clearing and re-running audio analysis doesn't change the result, so this is reproducible with these specific video files. Is it possible to at least detect this case (where something is clearly off because analysis thinks that there's 0 dynamic range and peaks are all at digital silence) and override the levels to be 0 instead of set to max gain?

I'm currently running version 25.0.18 (64 bit), but this has been happening consistently with the same set of files since 24

[Matt]

Could you provide a copy of one of the files that gets the wrong value to matt at jriver?  Thanks!

[halluci293]

I'm not really sure I know a good way to share one of these files as they are many GB bluray rips.

However, I think I have a lead: the files that fail all have in common that they are MKV containers with A_DTS audio tracks at 48k sampling frequency, 6 channels, and 24 bit depth. I'm thinking something about the format (particularly the 24-bit depth) might not be read properly when analyzing the audio. Interestingly, if I transcode these files in JRiver to MKV H264/AAC 1080p, they transcode fine and the converted audio also scans with no issues. If I try to transcode *just* the audio to FLAC, it fails (I'm not sure how to actually see logs from that process).

[halluci293]

In any case, it's not a huge deal to me if replaygain/r128 don't work for some files/formats. The bigger problem is that analysis appears to fail in a way that leaves tag values at extreme settings.

[RoderickGI]

I was going to suggest trying the Convert Video to Audio function in MC, and sending the audio file. But that actually re-encodes the audio to a new format, so that probably wouldn't work.

Maybe another tool could just strip the audio out, as is. You could then test if that gave the same analysis results, and if so, send that.

[syndromeofadown]

Here is a post of mine from 2014
https://yabb.jriver.com/interact/index.php/topic,81983.msg619277.html#msg619277

MC is applying an incorrect value of 52.00 dB for 'Volume Level (Replay Gain)' for few of my videos.
I imagine the correct value should be around 5 dB like most of my other videos.
This unnecessary 50± dB increase in volume is unbearably loud and is far beyond what my stereo can handle.

Out of a fairly large collection it is happening with three dvd rips and one blu-ray rip.
When analyzing the dvd's MC only takes one second to finish.
When analyzing the blu-ray everything appears normal. It takes about 30 seconds or so and stops at 100%.

I suspect these rips are from discs with some sort of error on them (I have 50 or so rips that just won't analyze) and
I doubt I'm the only user with these sound bombs in their library.
This could give an audiophile the worst day of their life.

Maybe a solution is to have a maximum value of 20 dB for Volume Level or give a warning before allowing playback of files with large values.

Back in 2014 I uploaded a DVD for JRiver then after an MC update I no longer had any issues with DVDs. I never sent a copy of my blu-ray that was causing problems but still can if needed. It's something by Steven Wilson, and the rip is to a hardisk folder not mkv. Internet is an issue for me but I'm sure I can get a copy sent one way or another if needed.

[halluci293]

OK, I found a smallish reproduction case by stripping out the video track and extracting just the audio track in mkvtoolnix to an MKA file. E-mail incoming.

[Matt]

Thanks for the file.  Investigation underway.

[Matt]

Coming next build:
Fixed: Analyzing audio streams with a format change near the end of the file could result in bogus R128/ReplayGain data.

Thanks to Hendrik for his detective work