How to reduce size of moov atom of H.264 movies to improve streaming start for smartphones?
Asked Answered
R

2

9

We run a video service streaming movies to smartphones (iOS&Android). We are encoding in H.264+AAC and using the mp4 container. We have a problem that long movies (60 minutes+) take a very long time to start playing and have tracked this down to the large size of moov atom for these movies. For 110 minute movies the atom is as large as 4.2Mb which obviously takes a long time to download to a smart-phone over 3G!

Is there anyway to make the moov atom smaller? We can reduce it bit by dropping the audio sampling rate, but obviously anything below 22kHz would not really be acceptable.

We are using ffmpeg as the encoder, and MP4Box to move the metadata to the front of the file. Is there any way to get it to make a smaller moov? Any other encoders out there which make a smaller moov?

For example...

Big size (280 Mb, 1h 49min) streamable mp4 (h.264, AAC) file have a big header size (4.2 Mb). File was encoded by two pass ffmpeg and MP4Box for replacing metadata into beginning of the file:

/usr/bin/ffmpeg -i /var/lib/encoder/incoming/2388 -aspect 320:210 -threads 8 -vcodec libx264 -profile baseline -level 13 -flags +loop+mv4 -cmp 256 -partitions +parti4x4+parti8x8+partp4x4+partp8x8+partb8x8 -me_method hex -subq 7 -trellis 1 -refs 5 -bf 0 -me_range 16 -g 250 -keyint_min 25 -sc_threshold 40 -i_qfactor 0.71 -qmin 10 -qmax 51 -qdiff 4 -b:v 270k -maxrate 270k -bufsize 270k -g 30 -passlogfile /tmp/mediaservice/3100/video-IPH.ffmpeg -an -f rawvideo -pass 1 -y /dev/null

/usr/bin/ffmpeg -i /var/lib/encoder/incoming/2388 -aspect 320:210 -threads 8 -vcodec libx264 -profile baseline -level 13 -flags +loop+mv4 -cmp 256 -partitions +parti4x4+parti8x8+partp4x4+partp8x8+partb8x8 -me_method hex -subq 7 -trellis 1 -refs 5 -bf 0 -me_range 16 -g 250 -keyint_min 25 -sc_threshold 40 -i_qfactor 0.71 -qmin 10 -qmax 51 -qdiff 4 -b:v 270k -maxrate 270k -bufsize 270k -g 30 -passlogfile /tmp/mediaservice/3100/video-IPH.ffmpeg -acodec libfaac -ac 2 -b:a 32k -ar 44100 -f mp4 -pass 2 -y /var/lib/encoder/encoded/3100/video-IPH.mp4

/usr/bin/MP4Box -quiet -tmp /tmp/mediaservice/3100/ -inter 500 /var/lib/encoder/encoded/3100/video-IPH.mp4

Media info (audio sample rate = 44100):

General
Count                                    : 278
Count of stream of this kind             : 1
Kind of stream                           : General
Kind of stream                           : General
Stream identifier                        : 0
Count of video streams                   : 1
Count of audio streams                   : 1
Video_Format_List                        : AVC
Video_Format_WithHint_List               : AVC
Codecs Video                             : AVC
Audio_Format_List                        : AAC
Audio_Format_WithHint_List               : AAC
Audio codecs                             : AAC LC
Complete name                            : 1348645218_970458_2465.iph.mp4
File name                                : 1348645218_970458_2465.iph.mp4
File extension                           : mp4
Format                                   : MPEG-4
Format                                   : MPEG-4
Format/Extensions usually used           : mp4 m4v m4a m4b m4p 3gpp 3gp 3gpp2 3g2 k3g jpm jpx mqv ismv isma f4v
Commercial name                          : MPEG-4
Format profile                           : Base Media
Internet media type                      : video/mp4
Codec ID                                 : isom
Codec ID/Url                             : http://www.apple.com/quicktime/download/standalone.html
Codec                                    : MPEG-4
Codec                                    : MPEG-4
Codec/Extensions usually used            : mp4 m4v m4a m4b m4p 3gpp 3gp 3gpp2 3g2 k3g jpm jpx mqv ismv isma f4v
File size                                : 272703970
File size                                : 260 MiB
File size                                : 260 MiB
File size                                : 260 MiB
File size                                : 260 MiB
File size                                : 260.1 MiB
Duration                                 : 6556027
Duration                                 : 1h 49mn
Duration                                 : 1h 49mn 16s 27ms
Duration                                 : 1h 49mn
Duration                                 : 01:49:16.027
Overall bit rate                         : 332767
Overall bit rate                         : 333 Kbps
Stream size                              : 4230761
Stream size                              : 4.03 MiB (2%)
Stream size                              : 4 MiB
Stream size                              : 4.0 MiB
Stream size                              : 4.03 MiB
Stream size                              : 4.035 MiB
Stream size                              : 4.03 MiB (2%)
Proportion of this stream                : 0.01551
HeaderSize                               : 4230683
DataSize                                 : 268473217
FooterSize                               : 70
IsStreamable                             : Yes
File last modification date              : UTC 2012-09-26 12:38:19
File last modification date (local)      : 2012-09-26 21:38:19
Writing application                      : Lavf54.6.100

Video
Count                                    : 201
Count of stream of this kind             : 1
Kind of stream                           : Video
Kind of stream                           : Video
Stream identifier                        : 0
ID                                       : 1
ID                                       : 1
Format                                   : AVC
Format/Info                              : Advanced Video Codec
Format/Url                               : http://developers.videolan.org/x264.html
Commercial name                          : AVC
Format profile                           : [email protected]
Format settings                          : 5 Ref Frames
Format settings, CABAC                   : No
Format settings, CABAC                   : No
Format settings, ReFrames                : 5
Format settings, ReFrames                : 5 frames
Format settings, GOP                     : M=1, N=30
Internet media type                      : video/H264
Codec ID                                 : avc1
Codec ID/Info                            : Advanced Video Coding
Codec ID/Url                             : http://www.apple.com/quicktime/download/standalone.html
Codec                                    : AVC
Codec                                    : AVC
Codec/Family                             : AVC
Codec/Info                               : Advanced Video Codec
Codec/Url                                : http://developers.videolan.org/x264.html
Codec/CC                                 : avc1
Codec profile                            : [email protected]
Codec settings                           : 5 Ref Frames
Codec settings, CABAC                    : No
Codec_Settings_RefFrames                 : 5
Duration                                 : 6556017
Duration                                 : 01:49:16.017
Bit rate                                 : 270000
Bit rate                                 : 270 Kbps
Width                                    : 480
Width                                    : 480 pixels
Height                                   : 270
Height                                   : 270 pixels
Pixel aspect ratio                       : 1.000
Display aspect ratio                     : 1.778
Display aspect ratio                     : 16:9
Rotation                                 : 0.000
Frame rate mode                          : CFR
Frame rate mode                          : Constant
FrameRate_Mode_Original                  : VFR
Frame rate                               : 29.970
Frame rate                               : 29.970 fps
Frame count                              : 196484
Resolution                               : 8
Resolution                               : 8 bits
Colorimetry                              : 4:2:0
Color space                              : YUV
Chroma subsampling                       : 4:2:0
Bit depth                                : 8
Bit depth                                : 8 bits
Scan type                                : Progressive
Scan type                                : Progressive
Interlacement                            : PPF
Interlacement                            : Progressive
Bits/(Pixel*Frame)                       : 0.070
Stream size                              : 220159060
Stream size                              : 210 MiB (81%)
Stream size                              : 210 MiB
Stream size                              : 210 MiB
Stream size                              : 210 MiB
Stream size                              : 210.0 MiB
Stream size                              : 210 MiB (81%)
Proportion of this stream                : 0.80732
Writing library                          : x264 - core 125
Writing library                          : x264 core 125
Writing library/Name                     : x264
Writing library/Version                  : core 125
Encoding settings                        : cabac=0 / ref=5 / deblock=1:0:0 / analyse=0x1:0x131 / me=hex / subme=7 / psy=1 / psy_rd=1.00:0.00 / mixed_ref=1 / me_range=16 / chroma_me=1 / trellis=1 / 8x8dct=0 / cqm=0 / deadzone=21,11 / fast_pskip=1 / chroma_qp_offset=-2 / threads=8 / lookahead_threads=1 / sliced_threads=0 / nr=0 / decimate=1 / interlaced=0 / bluray_compat=0 / constrained_intra=0 / bframes=0 / weightp=0 / keyint=30 / keyint_min=16 / scenecut=40 / intra_refresh=0 / rc_lookahead=30 / rc=2pass / mbtree=1 / bitrate=270 / ratetol=1.0 / qcomp=0.60 / qpmin=10 / qpmax=51 / qpstep=4 / cplxblur=20.0 / qblur=0.5 / vbv_maxrate=270 / vbv_bufsize=270 / nal_hrd=none / ip_ratio=1.40 / aq=1:1.00
Tagged date                              : UTC 2012-09-25 07:21:37

Audio
Count                                    : 169
Count of stream of this kind             : 1
Kind of stream                           : Audio
Kind of stream                           : Audio
Stream identifier                        : 0
ID                                       : 2
ID                                       : 2
Format                                   : AAC
Format/Info                              : Advanced Audio Codec
Commercial name                          : AAC
Format profile                           : LC
Codec ID                                 : 40
Codec                                    : AAC LC
Codec                                    : AAC LC
Codec/Family                             : AAC
Codec/CC                                 : 40
Duration                                 : 6556027
Duration                                 : 1h 49mn
Duration                                 : 1h 49mn 16s 27ms
Duration                                 : 1h 49mn
Duration                                 : 01:49:16.027
Bit rate mode                            : VBR
Bit rate mode                            : Variable
Bit rate                                 : 58955
Bit rate                                 : 59.0 Kbps
Maximum bit rate                         : 270000
Maximum bit rate                         : 270 Kbps
Channel(s)                               : 2
Channel(s)                               : 2 channels
Channel positions                        : Front: L R
Channel positions                        : 2/0/0
Sampling rate                            : 44100
Sampling rate                            : 44.1 KHz
Samples count                            : 289120791
Compression mode                         : Lossy
Compression mode                         : Lossy
Stream size                              : 48314149
Stream size                              : 46.1 MiB (18%)
Stream size                              : 46 MiB
Stream size                              : 46 MiB
Stream size                              : 46.1 MiB
Stream size                              : 46.08 MiB
Stream size                              : 46.1 MiB (18%)
Proportion of this stream                : 0.17717
Tagged date                              : UTC 2012-09-25 07:21:37

Moov atom info (/moov/trak[0] - video, /moov/trak[1] - audio) sample rate 44100: (look stsz and stts nodes in trak)

Atom ftyp @ 0 of size: 32, ends @ 32
Atom moov @ 32 of size: 4230651, ends @ 4230683
     Atom mvhd @ 40 of size: 108, ends @ 148
     Atom trak @ 148 of size: 868970, ends @ 869118
         Atom tkhd @ 156 of size: 92, ends @ 248
         Atom edts @ 248 of size: 36, ends @ 284
             Atom elst @ 256 of size: 28, ends @ 284
         Atom mdia @ 284 of size: 868834, ends @ 869118
             Atom mdhd @ 292 of size: 32, ends @ 324
             Atom hdlr @ 324 of size: 45, ends @ 369
             Atom minf @ 369 of size: 868749, ends @ 869118
                 Atom vmhd @ 377 of size: 20, ends @ 397
                 Atom dinf @ 397 of size: 36, ends @ 433
                     Atom dref @ 405 of size: 28, ends @ 433
                 Atom stbl @ 433 of size: 868685, ends @ 869118
                     Atom stsd @ 441 of size: 149, ends @ 590
                         Atom avc1 @ 457 of size: 133, ends @ 590
                             Atom avcC @ 543 of size: 47, ends @ 590
                     Atom stts @ 590 of size: 24, ends @ 614
                     Atom stss @ 614 of size: 26340, ends @ 26954
                     Atom stsc @ 26954 of size: 52, ends @ 27006
                     Atom stsz @ 27006 of size: 785956, ends @ 812962
                     Atom stco @ 812962 of size: 56156, ends @ 869118
     Atom trak @ 869118 of size: 3361468, ends @ 4230586
         Atom tkhd @ 869126 of size: 92, ends @ 869218
         Atom edts @ 869218 of size: 36, ends @ 869254
             Atom elst @ 869226 of size: 28, ends @ 869254
         Atom mdia @ 869254 of size: 3361332, ends @ 4230586
             Atom mdhd @ 869262 of size: 32, ends @ 869294
             Atom hdlr @ 869294 of size: 45, ends @ 869339
             Atom minf @ 869339 of size: 3361247, ends @ 4230586
                 Atom smhd @ 869347 of size: 16, ends @ 869363
                 Atom dinf @ 869363 of size: 36, ends @ 869399
                     Atom dref @ 869371 of size: 28, ends @ 869399
                 Atom stbl @ 869399 of size: 3361187, ends @ 4230586
                     Atom stsd @ 869407 of size: 91, ends @ 869498
                         Atom mp4a @ 869423 of size: 75, ends @ 869498
                             Atom esds @ 869459 of size: 39, ends @ 869498
                     **Atom stts @ 869498 of size: 2135816, ends @ 3005314**
                     Atom stsc @ 3005314 of size: 39712, ends @ 3045026
                     **Atom stsz @ 3045026 of size: 1129400, ends @ 4174426**
                     Atom stco @ 4174426 of size: 56160, ends @ 4230586
     Atom udta @ 4230586 of size: 97, ends @ 4230683
         Atom meta @ 4230594 of size: 89, ends @ 4230683
             Atom hdlr @ 4230606 of size: 33, ends @ 4230639
             Atom ilst @ 4230639 of size: 44, ends @ 4230683
                 Atom ©too @ 4230647 of size: 36, ends @ 4230683
                     Atom data @ 4230655 of size: 28, ends @ 4230683
Atom mdat @ 4230683 of size: 268473217, ends @ 272703900
Atom free @ 272703900 of size: 8, ends @ 272703908
Atom free @ 272703908 of size: 62, ends @ 272703970
------------------------------------------------------
Total size: 272703970 bytes; 50 atoms total. AtomicParsley version: 0.9.0 (utf8)
Media data: 268473217 bytes; 4230753 bytes all other atoms (1.551% atom overhead).
Total free atom space: 70 bytes; 0.000% waste. Padding available: 0 bytes.
------------------------------------------------------

After reencoding this movie with audio sample rate 11025 header size much less:

Media info (audio sample rate = 11025): (crop duplicate info)

General
***
HeaderSize                               : 1276359

Video
***

Audio
Count                                    : 169
Count of stream of this kind             : 1
Kind of stream                           : Audio
Kind of stream                           : Audio
Stream identifier                        : 0
ID                                       : 2
ID                                       : 2
Format                                   : AAC
Format/Info                              : Advanced Audio Codec
Commercial name                          : AAC
Format profile                           : LC
Codec ID                                 : 40
Codec                                    : AAC LC
Codec                                    : AAC LC
Codec/Family                             : AAC
Codec/CC                                 : 40
Duration                                 : 6556132
Duration                                 : 1h 49mn
Duration                                 : 1h 49mn 16s 132ms
Duration                                 : 1h 49mn
Duration                                 : 01:49:16.132
Bit rate mode                            : VBR
Bit rate mode                            : Variable
Bit rate                                 : 37991
Bit rate                                 : 38.0 Kbps
Maximum bit rate                         : 128000
Maximum bit rate                         : 128 Kbps
Channel(s)                               : 2
Channel(s)                               : 2 channels
Channel positions                        : Front: L R
Channel positions                        : 2/0/0
Sampling rate                            : 11025
Sampling rate                            : 11.025 KHz
Samples count                            : 72281355
Compression mode                         : Lossy
Compression mode                         : Lossy
Stream size                              : 31134257
Stream size                              : 29.7 MiB (12%)
Stream size                              : 30 MiB
Stream size                              : 30 MiB
Stream size                              : 29.7 MiB
Stream size                              : 29.69 MiB
Stream size                              : 29.7 MiB (12%)
Proportion of this stream                : 0.12327
Tagged date                              : UTC 2012-09-25 13:20:28

Moov atom info (/moov/trak[0] - video, /moov/trak[1] - audio) sample rate 11025:

Atom ftyp @ 0 of size: 32, ends @ 32
Atom moov @ 32 of size: 1276327, ends @ 1276359
     Atom mvhd @ 40 of size: 108, ends @ 148
     Atom trak @ 148 of size: 821662, ends @ 821810
         Atom tkhd @ 156 of size: 92, ends @ 248
         Atom edts @ 248 of size: 36, ends @ 284
             Atom elst @ 256 of size: 28, ends @ 284
         Atom mdia @ 284 of size: 821526, ends @ 821810
             Atom mdhd @ 292 of size: 32, ends @ 324
             Atom hdlr @ 324 of size: 45, ends @ 369
             Atom minf @ 369 of size: 821441, ends @ 821810
                 Atom vmhd @ 377 of size: 20, ends @ 397
                 Atom dinf @ 397 of size: 36, ends @ 433
                     Atom dref @ 405 of size: 28, ends @ 433
                 Atom stbl @ 433 of size: 821377, ends @ 821810
                     Atom stsd @ 441 of size: 149, ends @ 590
                         Atom avc1 @ 457 of size: 133, ends @ 590
                             Atom avcC @ 543 of size: 47, ends @ 590
                     Atom stts @ 590 of size: 24, ends @ 614
                     Atom stss @ 614 of size: 26340, ends @ 26954
                     Atom stsc @ 26954 of size: 52, ends @ 27006
                     Atom stsz @ 27006 of size: 785956, ends @ 812962
                     Atom stco @ 812962 of size: 8848, ends @ 821810
     Atom trak @ 821810 of size: 454452, ends @ 1276262
         Atom tkhd @ 821818 of size: 92, ends @ 821910
         Atom edts @ 821910 of size: 36, ends @ 821946
             Atom elst @ 821918 of size: 28, ends @ 821946
         Atom mdia @ 821946 of size: 454316, ends @ 1276262
             Atom mdhd @ 821954 of size: 32, ends @ 821986
             Atom hdlr @ 821986 of size: 45, ends @ 822031
             Atom minf @ 822031 of size: 454231, ends @ 1276262
                 Atom smhd @ 822039 of size: 16, ends @ 822055
                 Atom dinf @ 822055 of size: 36, ends @ 822091
                     Atom dref @ 822063 of size: 28, ends @ 822091
                 Atom stbl @ 822091 of size: 454171, ends @ 1276262
                     Atom stsd @ 822099 of size: 91, ends @ 822190
                         Atom mp4a @ 822115 of size: 75, ends @ 822190
                             Atom esds @ 822151 of size: 39, ends @ 822190
                     Atom stts @ 822190 of size: 161368, ends @ 983558
                     Atom stsc @ 983558 of size: 1480, ends @ 985038
                     Atom stsz @ 985038 of size: 282372, ends @ 1267410
                     Atom stco @ 1267410 of size: 8852, ends @ 1276262
     Atom udta @ 1276262 of size: 97, ends @ 1276359
         Atom meta @ 1276270 of size: 89, ends @ 1276359
             Atom hdlr @ 1276282 of size: 33, ends @ 1276315
             Atom ilst @ 1276315 of size: 44, ends @ 1276359
                 Atom ©too @ 1276323 of size: 36, ends @ 1276359
                     Atom data @ 1276331 of size: 28, ends @ 1276359
Atom mdat @ 1276359 of size: 251293325, ends @ 252569684
Atom free @ 252569684 of size: 8, ends @ 252569692
Atom free @ 252569692 of size: 62, ends @ 252569754
------------------------------------------------------
Total size: 252569754 bytes; 50 atoms total. AtomicParsley version: 0.9.0 (utf8)
Media data: 251293325 bytes; 1276429 bytes all other atoms (0.505% atom overhead).
Total free atom space: 70 bytes; 0.000% waste. Padding available: 0 bytes.
------------------------------------------------------

On slow connection this movie start playing after 30-40 seconds until header info (4.2 Mb) downloading. I need that movie start playing fast as it possible. And i have next questions:

  1. How reduce size of movie header?

  2. How reduce size of /moov[0]/trak[1]/mdia[0]/minf[0]/stbl[0] and why it so big when sample rate 44100?

Radio answered 27/9, 2012 at 8:26 Comment(3)
Did you solve your problem? I am now having the same issues and am struggling to find a solution. https://mcmap.net/q/1321193/-why-does-html5-video-with-very-large-h-264-encoded-mp4-with-faststart-ie-metadata-at-beginning-take-ages-to-load/1048589Utterance
No, i didn't. I switched to DASH & HLS protocols.Radio
ok not what I wanted to hear but thanks!Utterance
E
0

It looks like the AAC encoder or ffmpeg is doing a bad job encoding the AAC stream. The sample rate is not the issue here. Have you tried using the other AAC encoder?

-acodec aac -strict experimental

ffmpeg usually uses very small chunk sizes which will consequently lead to larger headers. Same goes for video.

The below is quite extreme.

Atom stts @ 869498 of size: 2135816, ends @ 3005314

Atom stsz @ 3045026 of size: 1129400, ends @ 4174426

I would try a different encoder if the other AAC encoder ends up looking the same. Instead of using ffmpeg and then mp4box, I recommend you to look at a branch of ffmpeg called ffmbc. This can put the header at the start while transcoding instead of it being a post process. But given it's an ffmpeg branch, I am not sure whether it will help on your header size issue. Worth giving a try though.

Ellingston answered 27/9, 2012 at 9:56 Comment(3)
I've reencoded movie ffmpeg -i 1348645218_970458_2465.iph.mp4 -acodec aac -strict experimental -vcodec copy 1348645218_970458_2465.iph.aac_exp.mp4 And header size reduce from 4.2 Mb to 3.8Mb. Any case it's big size. Maybe you know how increase chunk size?Radio
I am afraid that may not be possible using ffmpeg. It's really down to the encoder to do this. Try ffmbc too and see if that works better.Ellingston
This isn't an answer, but you can use ffmpeg to move the moov atom with the -faststart option (you'll need to compile ffmpeg for this option since it was introduced yesterday). FFmpeg also comes with the separate qt-faststart tool which can also relocate the atom.Frowzy
E
0

If you operate a video service and streaming single large files is obviously not the best way to do things. You need to convert/modify the moov atom heavily into fragments, fragments into small sections, possibly fragmented for 1-2 seconds, this way the video plays instantaneously without requiring to load long buffer time. This tells the encoder to start playing the video right after the moov loads.

This will get complicated at start if you have no idea how all of this works, you need to learn the art of MP4 atom box structure and it's ISO specifications. It's easy once fully understood.

Another easier option is to use some tools that helps to do this, such as transcode the large mp4 files into a fragmented mp4 file. Remember you need a fragmented mp4 video file and make sure the moov is always at the top of the mp4 atom box structure.

There is a lot of information on this, specially for video streaming. You can check this similar question (many have asked similar question): How to output fragmented mp4 with ffmpeg?

Exile answered 12/3, 2023 at 20:45 Comment(0)

© 2022 - 2024 — McMap. All rights reserved.