Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: File size and ms:ss ratio decrease when upgrading LAME (Read 2904 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

File size and ms:ss ratio decrease when upgrading LAME

I just upgraded from LAME 3.98.2 to 3.99.5. Using "-V 2 -q 2" I compared encodes from the old and new versions. Files from the new version were smaller. They also had less ms (mid-side stereo) frames, which could mean less efficient stereo compression due to encoding two separate channels and not taking advantage of things in common.

Is there a decrease in audio quality, or has LAME improved so much that it can maintain or improve quality despite this? Should I be using a particular recommended LAME version?

File size and ms:ss ratio decrease when upgrading LAME

Reply #1
Perhaps you shouldn’t be so quick to assume that these phenomena have negative implications.

Reduced file size could simply mean that LAME is confident that it can reproduce the same signal transparently with fewer bits. An increased number of non-joint frames could mean that LAME is being more careful. And so on.

AFAIK, aside from my usual advice that developers tend to know what they’re doing and are moving the version numbers up for a reason, there have not being any hugely significant or across-the-board regressions in 3.99, so it seems to represent fair progress; of course, I welcome any correction if I’m mistaken.

File size and ms:ss ratio decrease when upgrading LAME

Reply #2
See the following thread, which asks pretty much the same question:

http://www.hydrogenaudio.org/forums/index....showtopic=94058

See my reply (post #4) for test results comparing 3.98.4 to 3.99.5 at different VBR quality levels for a set of about 40 test files. You'll see that results for -V2 are mostly smaller, but it's not always the case.

File size and ms:ss ratio decrease when upgrading LAME

Reply #3
To the OP, that thread mentioned above, saying you're not the first to wonder, importanly points out that SFB21 Bloat has been greatly reduced by some clever work in the new version to better get around that VBR efficiency flaw in the MP3 specification that requires the first 16kHz of the spectrum to require far more bits than would otherwise be necessary if the 16kHz+ portion is to be encoded with efficiency. This is essentially a free lunch, by removing some of the wasted bitrate yet hit the same quality targets.

This won't affect -V3 to -V9 which use the -Y switch implicitly to avoid the associated bitrate bloat, but -V0 to -V2 should on average see bitrate reductions in this version of LAME for that reason, regardless of any psymodel changes.*

If you compare -V2 -Y and plain -V2 in the 3.98.2 version to -V2 -Y and plain -V2 in this 3.99.5 version, you may well notice that -V2 -Y is much closer between the two versions, reflecting relatively small psymodel changes between the two. Having only a minor difference in -V2 -Y bitrate, this would probably generate less surprise and less suspicion of a difference is quality.

(* Actually, if I recall correctly, because -V0 and -V1 are not just transparent but "transparent-plus-a-safety-margin", the LAME devs increased that safety margin a little using the savings of sorting out SFB21 bloat, so those bitrates are about the same as before. -V2 conversely is the point at which LAME has historically been tuned for transparency ever since it was known as --alt-preset standard and -V0 was formerly --alt-preset extreme)
Dynamic – the artist formerly known as DickD

File size and ms:ss ratio decrease when upgrading LAME

Reply #4
Thanks for the replies! It's nice to know that the size decrease is due to an improvement. I will continue using 3.99.5 and not worry about the differences I observed.