Trimming AAC non-destructive and pricisely

[passlogin]

Greetings, i've some dilemma here.
I'm trying to get the audio from a movie part (e.g. the opening, dialogues, etc.)
The problem is, mp3DirectCut, which the program i use to trim the unwanted parts, seems to display file length incorrectly.
WMP with LAV codec, displays the length properly. (for example the movie is 24 minutes long, mp3DirectCut displays the length is 25 minutes)
However, WinAmp also displays the file length incorrectly, 25 minutes.

What have i done is demux mkv with MKVcleaver, and open the AAC on mp3DirectCut.
FYI this is the properties

Audio
Format                                   : AAC
Format/Info                              : Advanced Audio Codec
Format version                           : Version 4
Format profile                           : LC
Bit rate mode                            : Variable
Channel(s)                               : 2 channels
Channel positions                        : Front: L R
Sampling rate                            : 48.0 KHz
Compression mode                         : Lossy
Stream size                              : 28.1 MiB (100%)

[Cornucopia]

Length is displaying incorrectly because the file is VBR and the app(s) don't know how to correctly calculate length for files of that type. CBR is usually "filesize = bitrate * running time", with running time being listed as # of samples or # of "frames/packets". You can't do that with VBR.
Not surprised. VBR audio is OFTEN a bitch to work with.

Can mp3directcut make the edits the way you'd like? If so, don't sweat it.

Scott

[passlogin]

Any app you know to trim AAC without reencoding?

[passlogin]

Length is displaying incorrectly because the file is VBR and the app(s) don't know how to correctly calculate length for files of that type. CBR is usually "filesize = bitrate * running time", with running time being listed as # of samples or # of "frames/packets". You can't do that with VBR.
Not surprised. VBR audio is OFTEN a bitch to work with.

Can mp3directcut make the edits the way you'd like? If so, don't sweat it.

Scott

Unfortunatle, for me, not. Because i want it to be precise

[hello_hello]

Does "Use TOC Correction for VBR" under the Settings menu make a difference if it's checked?

[passlogin]

It seems that option do nothing to the file length.
fileforum.betanews.com/review/1010041409/1/view < proof

[poisondeathray]

You can cut (copy) it accurately with ffmpeg . Don't demux from the original mkv container . Some mkv's have variable frame rate, and that can cause some programs to give you the running time discrepancies . If you leave it in the mkv, this will keep the timecodes.

use -ss for start time (in hh:mm: ss.ms notation) , and -t for duration

e.g. copy audio from 10:00 minutes start time , for 1 minute , 5 seconds duration (until 11 minute , 5 second , playing time)

ffmpeg -i input.mkv -vn -acodec copy -ss 00:10:00 -t 00:01:05 output.mp4

I left it as "mp4" container in this example but you can use raw aac e.g. output.aac

[passlogin]

You can cut (copy) it accurately with ffmpeg . Don't demux from the original mkv container . Some mkv's have variable frame rate, and that can cause some programs to give you the running time discrepancies . If you leave it in the mkv, this will keep the timecodes.

use -ss for start time (in hh:mm: ss.ms notation) , and -t for duration

e.g. copy audio from 10:00 minutes start time , for 1 minute , 5 seconds duration (until 11 minute , 5 second , playing time)

ffmpeg -i input.mkv -vn -acodec copy -ss 00:10:00 -t 00:01:05 output.mp4

I left it as "mp4" container in this example but you can use raw aac e.g. output.aac

After that, then demux to get the raw aac?
Does ffmpeg cut in 1 ms accuracy?

[poisondeathray]

You don't have to demux it in a separate step; just change it to "output.aac" instead of "output.mp4"

I'm not sure how accurate ; it's within a few ms for audio.

But video , it isn't accurate for "cutting" - only to nearest keyframe

[passlogin]

Is there any logical method out there?
Like BeSplit? Does it have proper VBR support?

[passlogin]

Seems the current best option are ffmpeg, but it is not accurate though. Few ~40ms.
Audacity reports ffmpeg output are slightly longer, yet no waveform difference as far as my eyes can tell.

[poisondeathray]

Did you specify the precise .ms in ffmpeg ? hh:mm: ss.ms , and it was still 40ms off ?

You can also try cutting with mkvtoolnix with audio only ie. .mka using timecodes (with video, you can only cut to nearest keyframe, but with audio it should be much more precise. If the video was I-frame only, you can cut with frame accuracy) ,then demux it with mkvextract

[passlogin]

Yes it is I-frame, all of them, in VirtualDub, all frames are marked [k].
It was not 40ms off, ffmpeg appends some silence at the end..

EDIT: Console output

PS C:\Users\User\Documents\ffmpeg-20130606-git--win64-static\bin> .\ffmpeg -i '..\..\..\videos\Angel Beats!\[SS]_Angel
_Beats!_-_03_(1280x720_H.264)_[7378C000].mkv' -ss 12:12.690 -to 12:18.946 -acodec copy ..\..\output1.aac
ffmpeg version N-53868-gc51654f Copyright (c) 2000-2013 the FFmpeg developers
built on Jun 6 2013 03:06:23 with gcc 4.7.3 (GCC)
configuration: --enable-gpl --enable-version3 --disable-w32threads --enable-avisynth --enable-bzlib --enable-fontconfi
g --enable-frei0r --enable-gnutls --enable-iconv --enable-libass --enable-libbluray --enable-libcaca --enable-libfreetyp
e --enable-libgsm --enable-libilbc --enable-libmodplug --enable-libmp3lame --enable-libopencore-amrnb --enable-libopenco
re-amrwb --enable-libopenjpeg --enable-libopus --enable-librtmp --enable-libschroedinger --enable-libsoxr --enable-libsp
eex --enable-libtheora --enable-libtwolame --enable-libvo-aacenc --enable-libvo-amrwbenc --enable-libvorbis --enable-lib
vpx --enable-libx264 --enable-libxavs --enable-libxvid --enable-zlib
libavutil 52. 35.100 / 52. 35.100
libavcodec 55. 15.100 / 55. 15.100
libavformat 55. 8.102 / 55. 8.102
libavdevice 55. 2.100 / 55. 2.100
libavfilter 3. 75.101 / 3. 75.101
libswscale 2. 3.100 / 2. 3.100
libswresample 0. 17.102 / 0. 17.102
libpostproc 52. 3.100 / 52. 3.100
[matroska,webm @ 000000000249a440] Unknown entry 0x437E
Last message repeated 4 times
Input #0, matroska,webm, from '..\..\..\videos\Angel Beats!\[SS]_Angel_Beats!_-_03_(1280x720_H.264)_[7378C000].mkv':
Metadata:
title : Angel Beats! - Episode 03 - My Song
creation_time : 2010-04-29 05:29:20
Duration: 00:24:29.93, start: 0.000000, bitrate: 2144 kb/s
Chapter #0.0: start 0.000000, end 89.965000
Metadata:
title : Opening
Chapter #0.1: start 90.005000, end 735.944000
Metadata:
title : Part A
Chapter #0.2: start 735.984000, end 1348.556000
Metadata:
title : Part B
Chapter #0.3: start 1348.596000, end 1454.912000
Metadata:
title : Ending
Chapter #0.4: start 1454.952000, end 1469.885000
Metadata:
title : Preview
Stream #0:0(jpn): Video: h264 (High), yuv420p, 1280x720, SAR 185:187 DAR 2960:1683, 23.98 fps, 23.98 tbr, 1k tbn, 47
.95 tbc
Stream #0:1(jpn): Audio: aac, 48000 Hz, stereo, fltp (default)
Stream #0:2: Subtitle: ssa (default)
Metadata:
title : Character Colour Style
Stream #0:3(eng): Subtitle: ssa
Metadata:
title : Plain Style
Codec 0x18000 is not in the full list.
Stream #0:4: Attachment: unknown_codec
Metadata:
filename : Dutch 766 Bold BT.ttf
mimetype : application/x-truetype-font
Codec 0x18000 is not in the full list.
Stream #0:5: Attachment: unknown_codec
Metadata:
filename : Hamilton-Medium.otf
mimetype : application/x-truetype-font
Codec 0x18000 is not in the full list.
Stream #0:6: Attachment: unknown_codec
Metadata:
filename : ShinnBold.ttf
mimetype : application/x-truetype-font
Codec 0x18000 is not in the full list.
Stream #0:7: Attachment: unknown_codec
Metadata:
filename : ShinnBoldItalic.ttf
mimetype : application/x-truetype-font
Codec 0x18000 is not in the full list.
Stream #0:8: Attachment: unknown_codec
Metadata:
filename : ShinnDemiBold.ttf
mimetype : application/x-truetype-font
Codec 0x18000 is not in the full list.
Stream #0:9: Attachment: unknown_codec
Metadata:
filename : ShinnMedium.ttf
mimetype : application/x-truetype-font
Codec 0x18000 is not in the full list.
Stream #0:10: Attachment: unknown_codec
Metadata:
filename : ShinnMediumItalic.ttf
mimetype : application/x-truetype-font
Output #0, adts, to '..\..\output1.aac':
Metadata:
title : Angel Beats! - Episode 03 - My Song
encoder : Lavf55.8.102
Chapter #0.0: start 0.000000, end 3.254000
Metadata:
title : Part A
Chapter #0.1: start 3.294000, end 6.256000
Metadata:
title : Part B
Stream #0:0(jpn): Audio: aac, 48000 Hz, stereo (default)
Stream mapping:
Stream #0:1 -> #0:0 (copy)
Press [q] to stop, [?] for help
size= 97kB time=00:00:06.27 bitrate= 126.7kbits/s
video:0kB audio:95kB subtitle:0 global headers:0kB muxing overhead 2.114934%

The intended length is 6256ms.

[poisondeathray]

Yes it is I-frame, all of them, in VirtualDub, all frames are marked [k].
It was not 40ms off, ffmpeg appends some silence at the end..

[SS]_Angel_Beats!_-_03_(1280x720_H.264)_[7378C000].mkv

Very unlikely this mkv is all I-frame (encoded with intra) , I'm 99.99% certain

The reason is you're decompressing the video in vdub (uncompressed video), because of the input filter used by vdub

Anyways, if you know the timecodes , you can remux a .mka (audio only mkv) , and try mkvtoolnix

[passlogin]

I don't know if external url is allowed, but this: http://forums.animesuki.com/showthread.php?t=34738 , explained why this job comes to incredibly hard.

[passlogin]

Yes it is I-frame, all of them, in VirtualDub, all frames are marked [k].
It was not 40ms off, ffmpeg appends some silence at the end..

[SS]_Angel_Beats!_-_03_(1280x720_H.264)_[7378C000].mkv

Very unlikely this mkv is all I-frame (encoded with intra) , I'm 99.99% certain

The reason is you're decompressing the video in vdub (uncompressed video), because of the input filter used by vdub

Anyways, if you know the timecodes , you can remux a .mka (audio only mkv) , and try mkvtoolnix

The video is CFR, yet the audio is VBR [edited to prevent misunderstanding]

[poisondeathray]

I don't know if external url is allowed, but this: http://forums.animesuki.com/showthread.php?t=34738 , explained why this job comes to incredibly hard.

Video is VFR, but audio is CFR (audio is always CFR) .

The video is CBR, yet the audio is VBR

VFR, CFR, VBR, CBR are all different things

Frame rate and Bit Rate are 2 different things

When you demux into a mka, it should all be CFR timecodes (no more VFR).

[El Heggunte]

I don't know if external url is allowed, but this: http://forums.animesuki.com/showthread.php?t=34738 , explained why this job comes to incredibly hard.

Video is VFR, but audio is CFR (audio is always CFR).

Not very-true. Compressed audio MAY use Variable Frame Duration, though MOST formats wisely don't do that.
WMA may have Variable Frame Duration, and Vorbis always has Variable Frame Duration. BTW, several lossless audio formats use VFD as well.

[poisondeathray]

I don't know if external url is allowed, but this: http://forums.animesuki.com/showthread.php?t=34738 , explained why this job comes to incredibly hard.

Video is VFR, but audio is CFR (audio is always CFR).

Not very-true. Compressed audio MAY use Variable Frame Duration, though MOST formats wisely don't do that.
WMA may have Variable Frame Duration, and Vorbis always has Variable Frame Duration. BTW, several lossless audio formats use VFD as well.

Technically , audio doesn't have "frame duration", but I guess you mean variable sample rate ? I remember some posts about WMA variable sample rates

I guess a better way of saying it is "VFR" refers to video only, not audio. The timecodes govern frame display time for the video, not audio playback .

[passlogin]

But, strangely, why audio and video have different timecodes? My guess are they run at different speed
Sorry about that: the audio is VBR
Wait, wait. Why audio's timecode had more frames defined that its video counterpart?

Video:

# timecodes format v1
Assume 29.970
1,35243,23.976

Audio:

# timecodes format v1
Assume 29.970
1,68903,46.953

[poisondeathray]

But, strangely, why audio and video have different timecodes? My guess are they run at different speed
Sorry about that: the audio is VBR
Wait, wait. Why audio's timecode had more frames defined that its video counterpart?

Video:

# timecodes format v1
Assume 29.970
1,35243,23.976

Audio:

# timecodes format v1
Assume 29.970
1,68903,46.953

VBR audio is difficult to edit precisely and can often be problematic, this was mentioned earlier by Scott

These timecodes are for CFR video . Notice there are no changes , every frame is the same . VFR video would have entries representing where the frame rate changes - again, that's the VIDEO frame rate

Audio doesn't use v1 timecodes like this (it's only for video), the reason it's different is just assuming a different timebase (notice it's double the fps, but double the number of frames)

[El Heggunte]

Technically , audio doesn't have "frame duration", but I guess you mean variable sample rate ? I remember some posts about WMA variable sample rates

Well, just keep in mind that a video frame is a rectangle between two video timestamps , whereas an audio frame is a small set of audio samples.

Variable playback-rate happens when an evil audio format is either wrapped in an inadequate container (AVI, which doesn't have such thing as audio timestamps), OR is parsed in a way that ignores the audio timestamps.

http://forum.doom9.org/showthread.php?p=1513679#post1513679

Any app you know to trim AAC without reencoding?

eac3to. BUT 1) it's a command-line application and 2) you must know what you're trying to achieve.

[passlogin]

Thanks for your advice.
Then if audio don't use timecodes, what will it use?
Oh yeah, audio frame? never heard that before.

[passlogin]

Yes it is I-frame, all of them, in VirtualDub, all frames are marked [k].
It was not 40ms off, ffmpeg appends some silence at the end..

[SS]_Angel_Beats!_-_03_(1280x720_H.264)_[7378C000].mkv

Very unlikely this mkv is all I-frame (encoded with intra) , I'm 99.99% certain

The reason is you're decompressing the video in vdub (uncompressed video), because of the input filter used by vdub

Anyways, if you know the timecodes , you can remux a .mka (audio only mkv) , and try mkvtoolnix

If i know the timecodes, how do i?
I prefer frame accuracy instead of timecodes.
I've found it on global tab of mkvmerge, is it right?

[smrpix]

Then if audio don't use timecodes, what will it use?

Samples/Sec. This is how the pros "cheat" 23.976 to 24, by modifying the expected sample rate.

[Cornucopia]

It's also known as pull-down (not to be confused with Telecine's Pull-down: 24->23.976 is a pull-down of 0.1% and 23.976/24->29.97/30 is a pull-down of 20%).

Scott

[passlogin]

Seems it better to remux with mkvmerge, since the difference between intended length with the output length is just 13ms