Intensity Stereo for low bitrates

classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

Intensity Stereo for low bitrates

Georg Holzmann
Dear list!

I would be interested to try intensity stereo in LAME when using low
bitrates.
This should be an advantage for e.g. podcasts, which can be hours long
and where it is sufficient to use lower bitrates (96k - 64k).
(more about my motivation in [1])

I did find the following (closed) ticket about intensity stereo:
http://sourceforge.net/p/lame/feature-requests/45/
And the the branch "takehiro-2002_05_07-experimental" in CVS, with some
commit messages which include changes for IS (see [2]).

I tried to backport the relevant changes from
"takehiro-2002_05_07-experimental" into the main branch, but so far I
did not succeed because the code bases diverged already a lot.

So my questions:
- Is there any (active) plan to support intensity stereo?
- Is it likely that takehiro-2002_05_07-experimental will ever get into
the main branch?

If someone has any hints or suggestions, how it would be easier to merge
these changes back, it would be highly appreciated!

Thanks for any feedback,
LG
Georg


[1]: Why MP3 Intensity Stereo?
Podcasts/radio are most of the time speech-only and can be very long,
therefore it is obviously important to create as small files as possible.
Common bitrates are 64k mono, 96k stereo (or lower).
More recent formats like OPUS or AAC are of course much better, but MP3
is the de-facto standard in podcasting distribution, mainly because of
the wide software and hardware support.
Another point is, that AFAIK in 2017 the remaining patent licenses
should expire and then it will be an even nicer distribution format. I
have no doubt that it will stay for a very long time.


[2]: Commit message which include intensity stereo in branch
"takehiro-2002_05_07-experimental":
- prepare for intensity-stereo <2002-11-10 Son>
- initial support of intensity stereo for lower bitrate it will be
enabled when 64-112kbps by default <2003-05-18 Son>
- fix low bitrate and intensity stereo <2003-10-30 Don>
- better intensity stereo band selection <2003-12-29 Mon>


------------------------------------------------------------------------------

_______________________________________________
Lame-dev mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/lame-dev
Reply | Threaded
Open this post in threaded view
|

Re: Intensity Stereo for low bitrates

FPatryk
As far as i know, the intensity stereo has been abandoned here since now
the algorithm of selecting between joint stereo MS and stereo is more
advanced than in FHG.


I see no need to re-enable it, (especially by default) because the
coding technologies have evolved much from that time.
Greetings, Patryk.



__________ Informacja programu ESET NOD32 Antivirus, wersja bazy sygnatur wirusów 12444 (20151021) __________

Wiadomoœć została sprawdzona przez program ESET NOD32 Antivirus.

http://www.eset.pl lub http://www.eset.com 



------------------------------------------------------------------------------
_______________________________________________
Lame-dev mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/lame-dev
Reply | Threaded
Open this post in threaded view
|

Re: Intensity Stereo for low bitrates

Georg Holzmann
Hallo Patryk!

Thanks for your answer!

What I forgot to mention: I am trying to encode low-bitrate CBR files,
not VBR!
Of course I did not mean that IS should be enabled by default, I just
wanted to try it.
(maybe someone has still a working binary?)

I did not do any listening tests, but some people claim that intensity
stereo should be better for this special use case (also in lame/TODO in
the LAME CVS repository - see [1]).

Below are examples of 96kbps and 80kbps encoded stereo files with LAME
and Fraunhofer - see [2].

Is it possible to reduce the artifacts in these (and similar) examples?
The exact stereo information would not be that important (something like
Parametric Stereo in HE-AACv2 would be nice;) ), that's why I wanted to
try if intensity stereo would help.

Or does someone have any other hints how to improve encoding of
low-bitrate CBR MP3 files?

Thanks a lot,
LG
Georg


[1] from CVS/lame/TODO, line 127:
"Use intensity stereo. This is a must-have for low bitrates, but if the
algorythm is very good it could also be used in every case."


[2] low-bitrate CBR examples:

Original Source File (44.1kHz, stereo):
https://auphonic.com/media/audio-examples/tim.wav

Encoded in 96kbps CBR (lame3.99.5 and Fraunhofer/iTunes12.2.2.25):
https://auphonic.com/media/audio-examples/tim-lame-96k.mp3
https://auphonic.com/media/audio-examples/tim-itunes-96k.mp3

Encoded in 80kbps CBR:
https://auphonic.com/media/audio-examples/tim-lame-80k.mp3
https://auphonic.com/media/audio-examples/tim-itunes-80k.mp3

It is possible to hear artifacts especially from 18s to 28s.
The Fraunhofer encoded files contain a lot of musical noise.

I used LAME default settings to encode these files:
lame -h --cbr -b 96 tim.wav tim-lame-96k.mp3
lame -h --cbr -b 80 tim.wav tim-lame-80k.mp3



On 2015-10-21 20:54, FPatryk wrote:

> As far as i know, the intensity stereo has been abandoned here since now
> the algorithm of selecting between joint stereo MS and stereo is more
> advanced than in FHG.
>
>
> I see no need to re-enable it, (especially by default) because the
> coding technologies have evolved much from that time.
> Greetings, Patryk.
>
>
>
> __________ Informacja programu ESET NOD32 Antivirus, wersja bazy
> sygnatur wirusów 12444 (20151021) __________
>
> Wiadomoœć została sprawdzona przez program ESET NOD32 Antivirus.
>
> http://www.eset.pl lub http://www.eset.com
>


------------------------------------------------------------------------------

_______________________________________________
Lame-dev mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/lame-dev