ass subtitle are stretched

[fate]

I compiled ffmpeg with the option --enable-libass on debian sid. I can add ass subtitle with the command -vf "ass=subtitle.ass" but there is a little problem, the subtitles add with ffmpeg are stretched. I don't have this problem with mencoder and I think ffmpeg and mencoder use the same library libass.
In the image to the top you have the subtite load with vlc, with the good size, and to the bottom the subtitle add with ffmpeg.
I tried with several video, codec and subtitle and the result is always the same result.

[cehoyos]

Please add a ffmpeg command line that allows to reproduce the problem together with complete, uncut console output.

[fate]

The ffmpeg command :

ffmpeg -i raw/Rocket_Girls_Raw_04.mp4 -vf "ass=ass/Rocket_04.ass" -vcodec libx264 -crf 26 -acodec libfaac -ab 128k -threads 4 -t "00:00:10" Rocket_Girls_04_test_ffmpeg_ass.mp4 

The console output :

ffmpeg version N-38845-g463705b Copyright (c) 2000-2012 the FFmpeg developers
  built on Mar 17 2012 17:11:41 with gcc 4.6.3
  configuration: --enable-libmp3lame --enable-libxvid --enable-libvorbis --enable-gpl --enable-libfaac --enable-libtheora --enable-zlib --disable-shared --enable-libx264 --enable-libdirac --enable-nonfree --enable-version3 --enable-libschroedinger --enable-avfilter --enable-libspeex --enable-libopenjpeg --enable-libgsm --enable-postproc --enable-pthreads --enable-libopencore-amrnb --enable-libopencore-amrwb --enable-ffplay --enable-pthreads --enable-libass --prefix=/usr/local --enable-x11grab --enable-runtime-cpudetect --enable-bzlib --enable-libdc1394 --enable-libvpx
  libavutil      51. 42.100 / 51. 42.100
  libavcodec     54. 10.100 / 54. 10.100
  libavformat    54.  2.100 / 54.  2.100
  libavdevice    53.  4.100 / 53.  4.100
  libavfilter     2. 65.100 /  2. 65.100
  libswscale      2.  1.100 /  2.  1.100
  libswresample   0.  7.100 /  0.  7.100
  libpostproc    52.  0.100 / 52.  0.100
Input #0, mov,mp4,m4a,3gp,3g2,mj2, from 'raw/Rocket_Girls_Raw_04.mp4':
  Metadata:
    major_brand     : isom
    minor_version   : 512
    compatible_brands: isomiso2avc1mp41
    creation_time   : 1970-01-01 00:00:00
    encoder         : Lavf53.5.0
  Duration: 00:23:44.34, start: 0.000000, bitrate: 5652 kb/s
    Stream #0:0(eng): Video: h264 (High 4:4:4 Predictive) (avc1 / 0x31637661), yuv420p, 720x480 [SAR 186:157 DAR 279:157], 5425 kb/s, 23.98 fps, 23.98 tbr, 24k tbn, 47.95 tbc
    Metadata:
      creation_time   : 1970-01-01 00:00:00
      handler_name    : VideoHandler
    Stream #0:1(jpn): Audio: ac3 (ac-3 / 0x332D6361), 48000 Hz, stereo, s16, 224 kb/s
    Metadata:
      creation_time   : 1970-01-01 00:00:00
      handler_name    : 
[buffer @ 0x26c43a0] w:720 h:480 pixfmt:yuv420p tb:1/1000000 sar:186/157 sws_param:
[ass @ 0x265a280] Added subtitle file: 'ass/Rocket_04.ass' (2 styles, 2 events)
[buffersink @ 0x2659dc0] auto-inserting filter 'auto-inserted scale 0' between the filter 'Parsed_ass_0' and the filter 'out'                      
[ass @ 0x265a280] auto-inserting filter 'auto-inserted scale 1' between the filter 'src' and the filter 'Parsed_ass_0'
[scale @ 0x269c700] w:720 h:480 fmt:yuv420p -> w:720 h:480 fmt:rgb24 flags:0x4
[scale @ 0x26bf3a0] w:720 h:480 fmt:rgb24 -> w:720 h:480 fmt:yuv420p flags:0x4
[libx264 @ 0x26c3900] using SAR=186/157
[libx264 @ 0x26c3900] using cpu capabilities: MMX2 SSE2Fast SSSE3 FastShuffle SSE4.1 Cache64
[libx264 @ 0x26c3900] profile High, level 3.0
[libx264 @ 0x26c3900] 264 - core 119 r2106 07efeb4 - H.264/MPEG-4 AVC codec - Copyleft 2003-2011 - http://www.videolan.org/x264.html - options: cabac=1 ref=3 deblock=1:0:0 analyse=0x3:0x113 me=hex subme=7 psy=1 psy_rd=1.00:0.00 mixed_ref=1 me_range=16 chroma_me=1 trellis=1 8x8dct=1 cqm=0 deadzone=21,11 fast_pskip=1 chroma_qp_offset=-2 threads=4 sliced_threads=0 nr=0 decimate=1 interlaced=0 bluray_compat=0 constrained_intra=0 bframes=3 b_pyramid=2 b_adapt=1 b_bias=0 direct=1 weightb=1 open_gop=0 weightp=2 keyint=250 keyint_min=23 scenecut=40 intra_refresh=0 rc_lookahead=40 rc=crf mbtree=1 crf=26.0 qcomp=0.60 qpmin=0 qpmax=69 qpstep=4 ip_ratio=1.40 aq=1:1.00
Output #0, mp4, to 'Rocket_Girls_04_test_ffmpeg_ass.mp4':
  Metadata:
    major_brand     : isom
    minor_version   : 512
    compatible_brands: isomiso2avc1mp41
    creation_time   : 1970-01-01 00:00:00
    encoder         : Lavf54.2.100
    Stream #0:0(eng): Video: h264 (![0][0][0] / 0x0021), yuv420p, 720x480 [SAR 186:157 DAR 279:157], q=-1--1, 24k tbn, 23.98 tbc
    Metadata:
      creation_time   : 1970-01-01 00:00:00
      handler_name    : VideoHandler
    Stream #0:1(jpn): Audio: aac (@[0][0][0] / 0x0040), 48000 Hz, stereo, s16, 128 kb/s
    Metadata:
      creation_time   : 1970-01-01 00:00:00
      handler_name    : 
Stream mapping:
  Stream #0:0 -> #0:0 (h264 -> libx264)
  Stream #0:1 -> #0:1 (ac3 -> libfaac)
Press [q] to stop, [?] for help
DTS -1001, next:-41709 st:0 invalid droping
frame=  240 fps= 61 q=-1.0 Lsize=     379kB time=00:00:09.92 bitrate= 313.0kbits/s                                                                 
video:215kB audio:157kB global headers:0kB muxing overhead 2.011653%
[libx264 @ 0x26c3900] frame I:3     Avg QP:21.79  size: 23661
[libx264 @ 0x26c3900] frame P:118   Avg QP:21.29  size:  1024
[libx264 @ 0x26c3900] frame B:119   Avg QP:27.89  size:   230
[libx264 @ 0x26c3900] consecutive B-frames: 33.3%  0.8%  2.5% 63.3%
[libx264 @ 0x26c3900] mb I  I16..4: 15.5% 45.6% 38.9%
[libx264 @ 0x26c3900] mb P  I16..4:  1.1%  1.2%  0.7%  P16..4: 12.5%  1.9%  0.9%  0.0%  0.0%    skip:81.9%
[libx264 @ 0x26c3900] mb B  I16..4:  0.1%  0.0%  0.0%  B16..8: 16.1%  0.3%  0.0%  direct: 0.1%  skip:83.4%  L0:28.3% L1:71.3% BI: 0.4%
[libx264 @ 0x26c3900] 8x8 transform intra:42.0% inter:80.8%
[libx264 @ 0x26c3900] coded y,uvDC,uvAC intra: 43.9% 40.6% 16.6% inter: 1.2% 2.9% 0.1%
[libx264 @ 0x26c3900] i16 v,h,dc,p: 26% 28%  8% 38%
[libx264 @ 0x26c3900] i8 v,h,dc,ddl,ddr,vr,hd,vl,hu: 22% 14% 26%  4%  6% 13%  4%  5%  5%
[libx264 @ 0x26c3900] i4 v,h,dc,ddl,ddr,vr,hd,vl,hu: 27% 15% 14%  6%  9% 12%  6%  8%  3%
[libx264 @ 0x26c3900] i8c dc,h,v,p: 62% 18% 17%  3%
[libx264 @ 0x26c3900] Weighted P-Frames: Y:0.0% UV:0.0%
[libx264 @ 0x26c3900] ref P L0: 74.9%  6.2% 15.4%  3.5%
[libx264 @ 0x26c3900] ref B L0: 90.9%  7.9%  1.2%
[libx264 @ 0x26c3900] ref B L1: 98.2%  1.8%
[libx264 @ 0x26c3900] kb/s:175.15

I attach the ass subtitle I used to made my tests.

[cehoyos]

Is the problem also reproducible without -vcodec libx264 (but with one of the internal encoders)?

[fate]

How I can use internal encoder ? (Sorry I'm a beginner on ffmpeg)
I tried with libxvid and I had the same problem.

[cehoyos]

$ ffmpeg -i raw/Rocket_Girls_Raw_04.mp4 -vf ass=ass/Rocket_04.ass -an -vcodec mpeg4 -qscale 2 -t 10 Rocket_Girls_04_test_ffmpeg_ass.mp4

[fate]

I have the same problem with your command.

[saste]

Confirm the problem:

ffplay -vf ass=/home/stefano/s/Rocket_04.ass -f lavfi testsrc=s=720x480

The problem depends on the fact that vf_ass.c uses the same DAR of input for rendering the subtitles, which is indeed a rather arbitrary behavior (see line in vf_ass.c: ass_set_aspect_ratio(ass->renderer, dar, sar)).

Possible fixes:
set the DAR to 1.0, or allow the user to specify the DAR through an option.

In the second case I ask: does it make sense to be able to set such value, or should be rather set in the ass file itself (assuming it is possible)?

[dbuitenh]

Replying to saste:

The problem depends on the fact that vf_ass.c uses the same DAR of input for rendering the subtitles, which is indeed a rather arbitrary behavior (see line in vf_ass.c: ass_set_aspect_ratio(ass->renderer, dar, sar)).

Possible fixes:
set the DAR to 1.0, or allow the user to specify the DAR through an option.

In the second case I ask: does it make sense to be able to set such value, or should be rather set in the ass file itself (assuming it is possible)?

The thing that everyone who uses VSFilter on windows (thus defining the behavior for other players and libs) has been doing since time immortal is to manually set the x/y resize factors in the .ass file. i.e. something with a 32:27 PAR -> x = (27 / 32) = 84.275%, y = 100 %. VSFilter then applies these at a 1.0 DAR, and then the renderer resizes on playback as the very last step.

To sum up: VSFilter always applies at 1.0 DAR.

[cehoyos]

Patch by Stefano that fixes the problem here:
 http://thread.gmane.org/gmane.comp.video.ffmpeg.devel/142213/focus=142325

[fate]

I managed to patch. It's work, a little too much same . Now the subtilte is little bit compress. You can see on the image. On the top it's the subtitle add with ffmpeg patched and on the bottom it's the subtitle load with vlc (it's the real size). But it's much better than before.
I will try to modify the patch, if I can, for adjust the size of subtitle.
Thank you :)

[saste]

Should be fixed in:

commit 247fbf071bb9d30f5341844a3d4ebbd53a986968
Author: Nicolas George <nicolas.george@normalesup.org>
Date:   Thu Mar 22 20:59:36 2012 +0100

    ass: fix aspect ratio computation.

Please reopen if it doesn't work for you.

[fate]

Ok I will try this patch. Thanks ;)