More > JRiver Media Center 19 for Windows
NEW: Improved audio analysis and volume leveling (R128)
mwillems:
--- Quote from: astromo on August 18, 2013, 04:48:12 am ---Currently reanalyzing my whole library, I'm super excited to try all this out.
Yep, I hear ya. I've got a separate machine set up just to do this one job. Audio only files didn't take too long. Analysing fine video's the "fun" job... ;)
--- End quote ---
Just finished up my re-analyzing last night, and I have to say, I'm pretty impressed so far. I cued up a play doctor playlist and listened for over an hour without having to adjust the volume once, which was never really the case with the old volume leveling (it would get close, but I'd find myself adjusting now and then even so). It's also neat to see all the new info about tracks. All in all a really neat improvement.
6233638:
--- Quote from: mwillems on August 18, 2013, 09:21:44 am ---Just finished up my re-analyzing last night, and I have to say, I'm pretty impressed so far. I cued up a play doctor playlist and listened for over an hour without having to adjust the volume once, which was never really the case with the old volume leveling (it would get close, but I'd find myself adjusting now and then even so). It's also neat to see all the new info about tracks. All in all a really neat improvement.
--- End quote ---
I agree - it's working really well when mixing random stereo tracks together.
When you introduce multichannel tracks - especially if you are downmixing to stereo - I think it could still use some work.
And album-based leveling now uses the gain from the loudest track rather than the average gain, which breaks leveling and has me changing the volume between albums again.
JustinM:
In the Past I never saw the value in this feature... Now that I've been listening to the r128 version.. Its a real treat not always playing with volume. It works great !
Congrats JRiver. an thanks for the feature I thought I didn't want :P
mwillems:
--- Quote from: 6233638 on August 18, 2013, 11:11:47 am ---I agree - it's working really well when mixing random stereo tracks together.
When you introduce multichannel tracks - especially if you are downmixing to stereo - I think it could still use some work.
And album-based leveling now uses the gain from the loudest track rather than the average gain, which breaks leveling and has me changing the volume between albums again.
--- End quote ---
I noticed the change in the way albums were handled this morning. I guess the idea is that it's designed to automatically prevent clipping if you use it with adaptive volume? Not sure why else it would change.
I think I agree with you though, while it works great for playlists, it does seem result in some pretty serious inter-album differences. I just tried a "torture test": switching from Slayer (loudest track requires -16.5 LU) to Berlioz (loudest track requires -2.3 LU) (a 14.2 LU spread in the volume leveling). To get those sounding about "the same" I had to make about a 3.5 dB adjustment, after volume leveling, to get subjectively similar sound (total of 17.7 difference). The averages for those albums would have been around -16 LU and +2.2 LU respectively (an 18.2 LU spread).
In this admittedly very small sample, the average would have covered my perceived difference in volume within a half dB, while the new "pick the lowest" method is much farther off. I'm sure it won't work out that way in every case, but I'd be curious to hear additional observations on the new album leveling method.
6233638:
--- Quote from: mwillems on August 18, 2013, 06:35:03 pm ---I noticed the change in the way albums were handled this morning. I guess the idea is that it's designed to automatically prevent clipping if you use it with adaptive volume? Not sure why else it would change.
--- End quote ---
Yes, I believe the change was made to avoid potentially clipping with album playback. But with analysis measuring the peak levels, it seems like it shouldn't be necessary.
--- Quote from: mwillems on August 18, 2013, 06:35:03 pm ---I think I agree with you though, while it works great for playlists, it does seem result in some pretty serious inter-album differences. I just tried a "torture test": switching from Slayer (loudest track requires -16.5 LU) to Berlioz (loudest track requires -2.3 LU) (a 14.2 LU spread in the volume leveling). To get those sounding about "the same" I had to make about a 3.5 dB adjustment, after volume leveling, to get subjectively similar sound (total of 17.7 difference). The averages for those albums would have been around -16 LU and +2.2 LU respectively (an 18.2 LU spread).
In this admittedly very small sample, the average would have covered my perceived difference in volume within a half dB, while the new "pick the lowest" method is much farther off. I'm sure it won't work out that way in every case, but I'd be curious to hear additional observations on the new album leveling method.
--- End quote ---
This has been my experience too - there's a noticeable difference between playlists of albums now that it's using the loudest track's gain rather than the average.
I did some analysis on my library a couple of weeks ago, and about a third of my albums had a 3dB or greater delta, and 10% had a 6dB or greater delta when using the loudest track compared to the average level.
That's not to say that using the average level will be perfect either, but perceptually it seemed to give much better results.
I'm not convinced that clipping is going to be a widespread issue, because most music seems to be fine when normalized to -23 LUFS.
I do think we need some way to control the amount of headroom that volume leveling has available though; 23dB is not enough for video leveling - it needs more like 30dB.
Linking the internal volume control to the target volume level, rather than applying volume after leveling has been performed seems like the neatest solution for this. (if you want -30 LUFS as a target, reduce the volume by 7dB)
Navigation
[0] Message Index
[#] Next page
[*] Previous page
Go to full version