MP3s have extra blank audio after being imported

Omsk411 · Post by **Omsk411** » Wed Dec 28, 2005 8:34 pm

For some reason, Live is adding a whole bunch of silence at the end of MP3s I've imported into my set. For example, a 7 minute song gains an additional 5 minutes of silence after importing. This only appears to happen to MP3s with VBR. CBR seems to decode properly.

Windows XP SP 2
Live 5.0.3
Virus TI Polar v1.0.8
M-Audio Trigger Finger
Dell Inspiron 5150 Laptop (3.06 GHz, 1GB RAM)

Post by **Alex** » Wed Dec 28, 2005 9:03 pm

Hi Omsk411,

this could be a problem how DirectShow decode MP3 files. You could try to install Quicktime which seems to behaves better concerning the decoding of MP3.
If Quicktime is installed Live is using it instead of DirectShow.

regards,
/Alex

Omsk411 · Post by **Omsk411** » Mon Jan 02, 2006 1:00 pm

Alex wrote:Hi Omsk411,

this could be a problem how DirectShow decode MP3 files. You could try to install Quicktime which seems to behaves better concerning the decoding of MP3.
If Quicktime is installed Live is using it instead of DirectShow.

regards,
/Alex

Hi Alex--QuickTime is installed. Got it as part of iTunes. Currently running version 7.0.3.

Ralf · Post by **Ralf** » Mon Jan 02, 2006 2:11 pm

Hi Omsk411,

could it be that your VBR mp3 files have filenames longer
than 63 characters? In this case QuickTime wouldn't be
able to open them.

Ralf

glannie · Post by **glannie** » Tue Jan 03, 2006 12:46 am

Im pretty sure this is the same problem Ive been having for ages also

Check this thread out Omsk411 to see if its the same issue... it might save Ralf and Alex some time
http://www.ableton.com/forum/viewtopic.php?t=28010

Ralf · Post by **Ralf** » Tue Jan 03, 2006 1:52 am

Well, maybe that is those obscure QuickTime Problem
a user had a while ago, where he could not open those
high bitrate vbr files in QuickTime, not even using
the QuickTime Player, and the error message was
very unspecific. Reinstalling QuickTime did not help,
but when he reinstalled Windows XP for other reasons,
suddenly QuickTime started to work.

So, Omsk411, could you please see if you can open
those files in Apples QuickTime player?

Not that I'm suggesting to reinstall Windows, but if that
is the problem again, I could try to get info from Apple
*while we have a system where the problem occurs*.

Thanks,
Ralf

PS. Omsk411, are you from Omsk, if that's not a secret?

difference · Post by **difference** » Tue Jan 03, 2006 2:31 am

I'm having this problem too. I'm wondering if it's a bit more widespread than is known, if people aren't realising. I'd thought nothing of the gap at the end of the waveform display until now today when I was looking at a problem where a couple of MP3s wouldn't decode at all.

( http://www.ableton.com/forum/viewtopic.php?t=31731 )

This is quite a big deal for me. A wav or aiff file is created in the decoding cache for every MP3 that is decoded so these extra minutes of silence are costing me 100s of Megs of HD Space!

I am running Live 5.03 on XP (SP1) on a seperate installation and really don't want to put QuickTime on there, especially as that doesn't seem to solve the problem.

Another question I have (as per the link above) is regarding the MP3s that won't decode at all. If WMP can read them on the same system why can't Live?

Ralf · Post by **Ralf** » Tue Jan 03, 2006 9:22 am

difference wrote:I'm having this problem too. I'm wondering if it's a bit more widespread than is known, if people aren't realising. I'd thought nothing of the gap at the end of the waveform display until now today when I was looking at a problem where a couple of MP3s wouldn't decode at all.

That DirectShow VBR problem is well-known and even described in the
help IIRC ("use QuickTime for VBR files"). Most of the time it is no
issue (besides the extra space requirements) because we compensate
for it by setting the clip/loop correctly.

( http://www.ableton.com/forum/viewtopic.php?t=31731 )

This is quite a big deal for me. A wav or aiff file is created in the decoding cache for every MP3 that is decoded so these extra minutes of silence are costing me 100s of Megs of HD Space!

I am running Live 5.03 on XP (SP1) on a seperate installation and really don't want to put QuickTime on there, especially as that doesn't seem to solve the problem.

100s of Megs, in the end, are about 3 to 10 decoded MP3 songs, and
they are cleaned up automatically. But if your harddisk is really small
(e.g. on a notebook) you should really consider installing QuickTime.
It installes itself into the task list at the lower right edge of the screen
(I hate that, personally), but you can open QuickTime settings, Extended
(or so) Tab, and turn that off, and there's a file type where you can
disable all types except QT movies, so QuickTime won't come in your way.

Another question I have (as per the link above) is regarding the MP3s that won't decode at all. If WMP can read them on the same system why can't Live?

It is a known fact that the WMP internally compensates for some problems
in the decoders which are by standard installed in Windows and used by
WMP. Maybe they are even using hidden decoders?
Probably that's the reason why Microsoft doesn't fix the problems.
So again, QuickTime significantly improves the situation.

Yours,
Ralf

difference · Post by **difference** » Tue Jan 03, 2006 7:01 pm

Thanks for the quick reply Ralf. I do dig the way you guys take the time to post on the forums.

Ralf wrote:
difference wrote:I'm having this problem too. I'm wondering if it's a bit more widespread than is known, if people aren't realising. I'd thought nothing of the gap at the end of the waveform display until now today when I was looking at a problem where a couple of MP3s wouldn't decode at all.
That DirectShow VBR problem is well-known and even described in the
help IIRC ("use QuickTime for VBR files"). Most of the time it is no
issue (besides the extra space requirements) because we compensate
for it by setting the clip/loop correctly.

Yeah, you're right - the extra space requirements is the only issue here. Every track I import is then warped and has loop markers etc set up ready for what I want to do with it anyway.

Ralf wrote:
( http://www.ableton.com/forum/viewtopic.php?t=31731 )

This is quite a big deal for me. A wav or aiff file is created in the decoding cache for every MP3 that is decoded so these extra minutes of silence are costing me 100s of Megs of HD Space!

I am running Live 5.03 on XP (SP1) on a seperate installation and really don't want to put QuickTime on there, especially as that doesn't seem to solve the problem.
100s of Megs, in the end, are about 3 to 10 decoded MP3 songs, and
they are cleaned up automatically. But if your harddisk is really small
(e.g. on a notebook) you should really consider installing QuickTime.
It installes itself into the task list at the lower right edge of the screen
(I hate that, personally), but you can open QuickTime settings, Extended
(or so) Tab, and turn that off, and there's a file type where you can
disable all types except QT movies, so QuickTime won't come in your way.

I like to move all my wavs into a seperate folder on my audio drive so I don't risk them being cleaned out of the decoding cache. I don't want to have to wait to decode a bunch of files again if I decide to bring in a load of tracks I haven't used for a while.

Regarding QuickTime, you can remove the icon but it will still be running in the background. I dunno, I just got the DAW OS stripped down to the bare minimum and I wanted to keep it that way really (doesn't matter for DJing but when I've got loads of VSTis and a huge Reason rack rewired I need every bit of CPU-juice I can squeeze out!)

Also, it seems QT is not solving the problem for everybody.

Ralf wrote:
Another question I have (as per the link above) is regarding the MP3s that won't decode at all. If WMP can read them on the same system why can't Live?
It is a known fact that the WMP internally compensates for some problems
in the decoders which are by standard installed in Windows and used by
WMP. Maybe they are even using hidden decoders?
Probably that's the reason why Microsoft doesn't fix the problems.
So again, QuickTime significantly improves the situation.

Playing devils advocate, could Live be programmed to do whatever WMP is doing (or perhaps have it's own decoding software) in the future? Maybe by licensing some codecs or something.

There are obviously work arounds and a bit of time and effort to trim the files and convert the (very occasional) one that won't decode isn't the end of the world. This is maybe a possible area for further development for you guys rather than a bug as such I suppose?

Anyway Ralf, thanks again for taking your time and thanks for what still is a great program.

bragi0 · Post by **bragi0** » Sat Jan 21, 2006 2:24 pm

OK, I worked around the bug (in the Fraunhofer mp3 DS filter) by changing to another DS MP3 filter.

I'm currently using MP3Source from http://www.dsp-worx.de/?n=9

I used http://www.free-codecs.com/DirectShow_F ... wnload.htm to change the merit on the two codecs. I made the Fraunhofer one Unlikely (0x00400000), and the FileSource(MP3) one Preferred (0x08000000).

I can confirm that Ableton is indeed using the codec I selected and that has fixed the issue. I'm going to test a couple more decoders for speed and quality. MADPlugin is next on the list

From my POV it's certainly better than having to install QT... I'm trying to keep this laptop as clean of everything as humanly possible (no A/V, no anti-spyware, only basic plugins, no other software than Live and this decoder). I'd been having some stability problems but that all seems to have cleared up nicely.

Oh, one big warning ahead of time... changing decoders (Fraunhofer/DS to QT, or MP3Source/DS, or MADPlugin, or MPG123/DS) can and most likely will move the sample offset. You'll find that all your markers will be correct relative to eachother, but offset incorrectly against the track. Simply select them all (click the first then shift-click the last), zoom right in on the first and correct its position. That'll be a total pain in the proverbial if you have a huge collection of tracks to do it against though.

Oh well, there's the next little project on the list... pull the .asd file format apart and write something that can play with it.

-update-

I'm now using the mpg123/mad plugin from http://f23.aaa.livedoor.jp/~kanetuki/download.html

Configured for MAD decoding I'm getting noticably better performance decoding mp3s, and subjectively better quality. Maybe when I have more time I'll run an mp3 through the different filters and see if there's any blind-test difference.

One of the nice things about Ableton using DS/QT for decoding is that we actually get the option of using different decoders if we were having issue. Big ups to the ableton devs

Ralf · Post by **Ralf** » Mon Jan 23, 2006 8:03 pm

Hello bragi0,

thanks for your valuable hints.

I've spent a couple of hours testing the various codecs concerning
the "offset from start" problem. It would really be a problem if switching
between standard codecs (for example, because you move from PC to Mac)
would change the timing, so I was a bit in panic first, but it appears to be
not that bad.

I've tested the QuickTime 7.0.4 codec and three DirectShow Codecs,
the one which comes with XP by default (l3codecx.ax) and two codecs
delivered with Windows Media Players 9 resp. 10 (l3codeca.acm and
l3codecp.acm), which we prepare to support with the next Live release
(bugfix or other, whatever the next will be) - if they work as expected.

They all are Fraunhofer codecs, and not surprisingly, they all generate
the same output, sample by sample. So for the user not setting up
other codecs everything is fine.

The MP3Source codec produces exactly the same output as the Fraunhofer
codecs, but with 50 ms more silence at the beginning (for a certain
example). Who knows where this comes from.

I've read the MPEG standard back and forth, it's really a hard reading,
there is no 100% clear statement that there should be no timing offset.
Although sentences like the following suggest that there should be none:

"The delay due to the implementation of the decoder must be accounted for when calculating the actual presentation time from the PTS. In normal operation (not trick modes) all presentation units in the program shall be presented, and the variation in the difference between the PTS and the actual presentation times of presentation units shall not be perceivable."

At least the test setup in the standard does not allow for a time offset
compensation (delay or so) before calculating the difference, so no delay
seems to be allowed. The difference btw. must be less than 2power-15 / sqrt(12) in RMS and less that 2power-14 in absolute values, i.e. almost null.

Didn't test the mpg123/MAD filter because it first needs to be built,
but it contains a filter which is on by default, so the result would be
different too for the filter latency.

So the bottom line is, other codecs may have advantages, e.g. a little
lower delay when prehearing, better vbr support without a need to install
QuickTime, but they don't differ in quality of the decoded signal (at least if
they comply to the standard), so use them only if the delay is no
problem for you (that's what you said essentially).

The audio signal offset seems to be no problem for the default codecs.

Yours,
Ralf