Kodi Community Forum

Full Version: How to boost center channel when downmixing
You're currently viewing a stripped down version of our content. View the full version with proper formatting.
Pages: 1 2 3 4 5 6 7
I've upgraded from Frodo to Gotham but now I can't find how to boost the center channel. I have my raspberry hooked to my tv (no avr) and the voices in movies sound very low compared to Frodo, where I could select "boost center channel when downmixing" and it worked really well. I have tried checking "Normalize levels on downmix" but I can't get it to work like it did in Frodo.

Can anyone help me?
(2014-05-28, 23:56)Fraludio Wrote: [ -> ]I've upgraded from Frodo to Gotham but now I can't find how to boost the center channel. I have my raspberry hooked to my tv (no avr) and the voices in movies sound very low compared to Frodo, where I could select "boost center channel when downmixing" and it worked really well. I have tried checking "Normalize levels on downmix" but I can't get it to work like it did in Frodo.

That was a custom patch for Pi on Frodo that I added. I don't believe I've ever seen anyone report that it was useful.
On Gotham we are now using more of the standard xbmc Audio Engine where that isn't supported.
You should now get the same audio on Pi as on other platforms.

If a reasonable number of people report they were using it successfully on Frodo, then I can investigate implementing it on Gotham.
I'm using it in four pi's (my house, my parents', my girlfriend's and her parents') and it makes a really good difference to the point I bought the pi just because of this. Today i upgraded to gotham and the first movie we tried (Thor 2) my girlfriend and me went "wtf, it sounds awful, the voices are barely audible".
I would second the need for this.

XBMC (all platforms) really does need a boost centre channel volume during downmix option. (as distinct from volume amplification, which is really just a gain increase for all channels with a peak limiter, which doesn't solve the problem of recessed dialogue in the centre channel on many files)

Either that or a way to adjust the down mix matrix from advancedsettings.xml, so those of us that care enough can do something to adjust it without having to recompile. Kind of strange that the audio engine was rewritten so entirely without any thought to making the down mix matrix user adjustable in any way.
I just had a little chat with FernetMenta, because I had the feeling that we do the boost of the center channel on downmix by default now and thus removed that setting. And he confirmed. So, we do it by default UNLESS you enabled "normalize levels on downmix" (which is to be expected). So disable normalization and you should be good to go again. Note that this requires more CPU on the PI because of additional checks for audio clipping.

edit: I dared to change this issue and mark it as solved. Feel free to change it back if it doesn't work for you
(2014-05-30, 11:32)da-anda Wrote: [ -> ]I just had a little chat with FernetMenta, because I had the feeling that we do the boost of the center channel on downmix by default now and thus removed that setting. And he confirmed. So, we do it by default UNLESS you enabled "normalize levels on downmix" (which is to be expected). So disable normalization and you should be good to go again. Note that this requires more CPU on the PI because of additional checks for audio clipping.
Interesting - do you know roughly when the behaviour changed ? I've run so many different versions, and debug builds in the run up to Gotham on both Mac and Pi that I've lost track of which ones I was having problems with the centre channel volume. (and of course not all video file exhibit the problem in an obvious way)

So you're saying when normalization is turned off (which is not the default) that as well as boosting (and compressing/limiting) all channels to some degree, some additional boost (how much ?) is given to the centre channel compared to the left, right and surround channels ?

Just trying to understand the situation a bit better. Smile
We use ffmpeg's default mixing matrix. Some channels which are in the input stream but not in the audio configuration get mixed into those available. i.e. when playing 5.1 on a 2.0 setup the center channel gets mixed 0.5 into FL and 0.5 into FR. Default is that we normalize levels on downmix because this protects from clipping and keeps the result closest to the original. Not applying normalization, formally known as "boost volume on downmix" tries to keep the overall volume of the source. In particular this is desired when watching tv where the stream often switches from 5.1 to 2.0 and back.
@da-anda @FernetMenta

On Frodo, where the Pi has it's own mixing matrix generation code I added a new settings option "boost centre channel on downmix".
This added 3dB to centre channel after the mixing matrix was calculated.
This tends to make dialogue more audible amongst loud sound effects/background music.

On Gotham I use exactly the same mixing matrix as xbmc does on other platforms and this feature has been dropped.

So no, "normalise on downmix" is not related to the option I added, and so this issue is not solved.
well then, this sounds like an option good enough for core and not just PI specific, or, if we can, we could just do it by default for video playback (not music downmix)
(2014-05-30, 13:05)da-anda Wrote: [ -> ]well then, this sounds like an option good enough for core and not just PI specific, or, if we can, we could just do it by default for video playback (not music downmix)

It only applied to multichannel sources which is fairly rare with music.

There is a DSP add-on coming. So far I don't think that supports adjusting the mixing matrix, but it could do.
However I suspect the architecture of the DSP add-on will mean a separate processing stage which will be computationally expensive,
rather than adjusting the mixing matrix which is free, so that may not be an option for Pi.
I agree multichannel music is rare but I've got some examples that play well on the Pi and if centre channel boosting is ever added to core it should take multichannel music into consideration (ie no boost).
(2014-05-30, 13:05)da-anda Wrote: [ -> ]well then, this sounds like an option good enough for core and not just PI specific, or, if we can, we could just do it by default for video playback (not music downmix)
I would love to see it in mainline XBMC for all platforms, as long as it's an option that can be turned on and off like it was on RPi Frodo. (And probably off by default to be conservative) Perhaps even with the default amount of boost (3dB ?) when enabled being possible to override in advancedsettings. (Somewhere between 3dB and 6dB should cover most cases, but a fixed value isn't very flexible)

It might seem like an insignificant feature, but when you sometimes either can't hear the dialogue or are being blasted by sound effects and having to constantly ride the volume control it really does spoil the enjoyment of trying to watch something...
(2014-05-30, 14:07)DBMandrake Wrote: [ -> ]It might seem like an insignificant feature, but when you sometimes either can't hear the dialogue or are being blasted by sound effects and having to constantly ride the volume control it really does spoil the enjoyment of trying to watch something...
True, it makes watching a movie (even with a 2.1 speaker set) a real pain in the *ss.
Come to think of it, rather than having an advancedsetting option to adjust the amount of boost for all videos and a global on/off in settings, it would make a lot of sense to add a "centre channel boost" (or similar) option in the onscreen audio controls which can be saved as a default and also adjusted individually for different movies, just like the volume amplification setting can be - because not all videos require the centre channel boosting. (This also solves the problem of not accidentally applying it to multichannel music files)

So basically like the volume amplification setting but ONLY for the centre channel, and applied on top of any other correction like volume amplification. The only problem I can see with implementing this is would it need skin support to add it to the OSD overlay UI ?
Any news?.
Pages: 1 2 3 4 5 6 7