IPB

Welcome Guest ( Log In | Register )

 
Reply to this topicStart new topic
Negative gains may result in slightly lower bitrates, 711 MiB gained over a total of 27 GiB
skamp
post Nov 28 2012, 21:36
Post #1





Group: Developer
Posts: 1343
Joined: 4-May 04
From: France
Member No.: 13875



I started experimenting a bit after Greynol said that he applies album gain prior to MP3 encoding. I did the same while transcoding my entire library to Vorbis (AoTuV) -q 5: the resulting files take 711 MiB less than their counterpart (gain unchanged), over a total of 27.2 GiB. That's a 2.56% gain; average bitrate went down from 161 kbps to 157 kbps.

Obviously, I got those results because most of my music gets negative Replaygain values. As far as I'm concerned, it's interesting because I'm pretty short on space on my Clip's 32 GB microsdhc card.


--------------------
caudec.net
Go to the top of the page
+Quote Post
greynol
post Nov 28 2012, 21:43
Post #2





Group: Super Moderator
Posts: 10000
Joined: 1-April 04
From: San Francisco
Member No.: 13167



Yes I also notice a slight decrease in average bitrate as a result of attenuating prior to encoding.

This is nothing new however:
http://www.hydrogenaudio.org/forums/index....c=10637&hl=

The OP in that link is the author of Omni Encoder which, IIRC can scale prior to encoding.
http://www.hydrogenaudio.org/forums/index....c=51093&hl=

This post has been edited by greynol: Nov 28 2012, 21:54


--------------------
Your eyes cannot hear.
Go to the top of the page
+Quote Post
skamp
post Nov 28 2012, 21:51
Post #3





Group: Developer
Posts: 1343
Joined: 4-May 04
From: France
Member No.: 13875



Thanks for that. Caudec in SVN can now do it too:
CODE
caudec -c ogg -q 5 -G album *.flac


--------------------
caudec.net
Go to the top of the page
+Quote Post
greynol
post Nov 28 2012, 21:56
Post #4





Group: Super Moderator
Posts: 10000
Joined: 1-April 04
From: San Francisco
Member No.: 13167



No problem. Another benefit of scaling downwards prior to encoding is that clipping during lossy decoding can be mitigated, for those who are worried about it.


--------------------
Your eyes cannot hear.
Go to the top of the page
+Quote Post
BFG
post Nov 28 2012, 22:05
Post #5





Group: Members
Posts: 205
Joined: 22-July 12
Member No.: 101637



I suppose that this makes sense, in that some of the quieter sections of the track will be attenuated down to (near) silence, so the lossy encoder only needs to approximate a narrower section of the audio spectrum.

EDIT: The topics that greynol links to seems to explain this quite well. Looks like I have some reading to do smile.gif

This post has been edited by BFG: Nov 28 2012, 22:07
Go to the top of the page
+Quote Post
Dynamic
post Nov 29 2012, 02:13
Post #6





Group: Members
Posts: 793
Joined: 17-September 06
Member No.: 35307



Aside from Absolute Threshold of Hearing (ATH) models (or adaptive ATH models) being able to shed some info, it wasn't all that obvious why most codecs gained much by reducing volume given that nothing gets lost as such - it's pretty much all ratios used in psychoacoustic decisions, and the scaling is done in 32-bit floating point in virtually any codec, and really only ATH is based on absolute rather than relative loudness.

In LAME MP3 (in that first thread APS was --alt-preset standard, the old version of what's now -V2) there was a lot to gain by overcoming the sfb21 limitation of mp3 which means that bits don't need to be wasted below 16kHz - and this was demonstrated by using the -Y switch in LAME which saw scaling save only 0.5 kbps versus about 20 kbps without -Y in one example.
Go to the top of the page
+Quote Post

Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

 



RSS Lo-Fi Version Time is now: 17th April 2014 - 09:27