October 2009 - ffmpeg-user - lists.ffmpeg.org

preserving tkhd transformation matrix through transcode in ffmpeg?
by sgendler＠ideasculptor.com 31 Oct '09

31 Oct '09

I am transcoding iphone video files in ffmpeg. I have no trouble doing so except that the transformation matrix in the tkhd node is both ignored by ffmpeg and it is not copied through to the resulting quicktime (mov) file. The end result is that once a video has been transcoded, all knowledge of its original orientation is lost. I'd be happy to have ffmpeg run the video content through the matrix before transcoding, resulting in a video that has been correctly rotated according to the iphone orientation. Id be just as happy to have the matrix copied straight through to the new mov file so that any player that correctly uses the matrix will render the transcoded file correctly. Currently I have neither, which means that video shot with the phone in portrait mode get rendered sideways by ALL players once they have been transcoded by ffmpeg. This creates a big problem when streaming a transcoded file back to the iphone, since the quicktime player built into the phone expects to use the matrix, so it plays all videos shot in portrait mode sideways, and no matter how you rotate the phone, the video rotates to be sideways. All my other playback occurs via flash, and I had to write code to extract the rotation from a video before transcode and then pass that to my player, which I can then cause to rotate the video according to the stored metadata, but this is far from optimal. Ideally, my transcoded files would be rotated during the transcode process by feeding each pixel through the matrix. Is there a solution to this problem that I am missing somewhere? I've searched all over the net and tested the very latest svn revisions of ffmpeg, to no avail. Here's the commandline I am using: ffmpeg -y -i IMG_0031.MOV -vpre ./libx264-vidme_video.ffpreset -ar 22050 -s 600x450 -f mov IMG_0031.ffmpeg.mov I've attached the preset file output is: FFmpeg version SVN-r20416, Copyright (c) 2000-2009 Fabrice Bellard, et al. built on Oct 29 2009 16:27:31 with gcc 4.1.2 20071124 (Red Hat 4.1.2-42) configuration: --prefix=/usr --enable-gpl --enable-postproc --enable-avfilter --enable-avfilter-lavf --enable-libmp3lame --enable-libvorbis --enable-libtheora --enable-libdirac --enable-libschroedinger --enable-libfaac --enable-libfaad --enable-libxvid --enable-libx264 --mandir=/usr/share/man --enable-pthreads --enable-libspeex --incdir=/usr/include --extra-cflags=-I/usr/include --enable-nonfree --extra-ldflags=-L/usr/lib --enable-libopencore-amrnb --enable-libopencore-amrwb --enable-version3 --enable-shared --disable-debug --enable-nonfree libavutil 50. 3. 0 / 50. 3. 0 libavcodec 52.37. 1 / 52.37. 1 libavformat 52.39. 2 / 52.39. 2 libavdevice 52. 2. 0 / 52. 2. 0 libavfilter 1. 6. 0 / 1. 6. 0 libswscale 0. 7. 1 / 0. 7. 1 libpostproc 51. 2. 0 / 51. 2. 0 Seems stream 1 codec frame rate differs from container frame rate: 1200.00 (1200/1) -> 30.00 (30/1) Input #0, mov,mp4,m4a,3gp,3g2,mj2, from 'IMG_0031.MOV': Duration: 00:00:33.11, start: 0.000000, bitrate: 824 kb/s Stream #0.0(und): Audio: aac, 44100 Hz, 1 channels, s16, 79 kb/s Stream #0.1(und): Video: h264, yuv420p, 480x360, 742 kb/s, 30 tbr, 600 tbn, 1200 tbc Metadata major_brand : qt minor_version : 0 compatible_brands: qt [libx264 @ 0x8094f20]using cpu capabilities: MMX2 SSE2Slow [libx264 @ 0x8094f20]profile Baseline, level 3.0 Output #0, mov, to 'IMG_0031.ffmpeg.mov': Stream #0.0(und): Video: libx264, yuv420p, 600x450, q=10-51, 500 kb/s, 30 tbn, 30 tbc Stream #0.1(und): Audio: aac, 22050 Hz, 1 channels, s16, 64 kb/s Stream mapping: Stream #0.1 -> #0.0 Stream #0.0 -> #0.1 Press [q] to stop encoding frame= 992 fps= 14 q=15181029.0 Lsize= 2149kB time=32.97 bitrate= 534.0kbits/s video:1981kB audio:153kB global headers:0kB muxing overhead 0.685415% [libx264 @ 0x8094f20]frame I:34 Avg QP:31.58 size: 6466 [libx264 @ 0x8094f20]frame P:958 Avg QP:30.22 size: 1888 [libx264 @ 0x8094f20]mb I I16..4: 74.4% 0.0% 25.6% [libx264 @ 0x8094f20]mb P I16..4: 11.1% 0.0% 1.6% P16..4: 29.5% 7.5% 0.8% 0.0% 0.0% skip:49.5% [libx264 @ 0x8094f20]final ratefactor: 29.26 [libx264 @ 0x8094f20]coded y,uvDC,uvAC intra: 20.4% 21.7% 0.7% inter: 5.3% 3.7% 0.0% [libx264 @ 0x8094f20]i16 v,h,dc,p: 59% 13% 17% 12% [libx264 @ 0x8094f20]i4 v,h,dc,ddl,ddr,vr,hd,vl,hu: 32% 14% 25% 5% 5% 6% 4% 6% 3% [libx264 @ 0x8094f20]kb/s:490.72

2 4

Problems creating high-quality VBR mjpeg files
by krueger＠signal7.de 30 Oct '09

30 Oct '09

Hi, I'm trying to transcode AVCHD material from a camera (Panasonic DMC- GH1) to mjpeg VBR files and I'm hitting a bitrate limit which I would like to exceed but can't seem to achieve that. What I tried was: ffmpeg -i file.mts -vcodec mjpeg -qscale 1 -acodec copy file.mov ffmpeg -i file.mts -vcodec mjpeg -b 60000k -acodec copy file.mov The input file is a 720p at 50 fps and the bitrates I get are approx. 20MBit/s which is roughly 50k per frame which is rather bad (I see a lot of artifacts compared to the source material). How can I force the bitrate to go higher while keeping it VBR? What I would like to tell ffmpeg is to do what it does now but multiply quality by factor X. I started adding a "*2" to the default rate control equation but there I really don't know what I am doing (and it didn't seem to change anything), so I thought I'd ask. Thanks in advance for any hints, Robert

3 8

why doe DeInterlace fail
by jshupert＠pps-inc.com 30 Oct '09

30 Oct '09

friends, I wish to make a 640x480 deinterlaced mp4 from a D1 (720x480) photo jpeg mov. I do make a file but the deinterlacing fails. I wonder what I have wrong? Why does Deinterlace fail? -here is my command ffmpeg -i /mnt/data_1/vid_1/pj2.mov -vcodec libxvid -b 900000 -acodec libfaac -deinterlace -ab 128000 -ac 2 -aspect 4:3 -r 30000/1001 -s 640x480 -async 1 -vsync 1 -y /mnt/data_1/vid_1/pj2-apprv.mp4 -here is a snippet of the return from my command -- snipped due to the extreme length of the many lines of "Deinterlacing failed" <snip> Deinterlacing failed Deinterlacing failed Deinterlacing failedq=6.0 size= 4233kB time=31.70 bitrate=1094.0kbits/s Deinterlacing failed frame= 956 fps= 14 q=6.0 Lsize= 4299kB time=31.85 bitrate=1105.8kbits/s video:3780kB audio:498kB global headers:0kB muxing overhead 0.482740% -here is my ffmpeg vers FFmpeg version SVN-r19890, Copyright (c) 2000-2009 Fabrice Bellard, et al. configuration: --enable-libmp3lame --enable-libtheora --enable-libvorbis --enable-libfaac --enable-libfaad --disable-mmx --enable-shared --enable-libxvid --enable-gpl --enable-nonfree libavutil 50. 3. 0 / 50. 3. 0 libavcodec 52.35. 0 / 52.35. 0 libavformat 52.38. 0 / 52.38. 0 libavdevice 52. 2. 0 / 52. 2. 0 libswscale 0. 7. 1 / 0. 7. 1 built on Sep 16 2009 18:42:46, gcc: 4.1.2 20080704 (Red Hat 4.1.2-44) thanks much! jS

1 0

memory usage of ffplay
by m.louvel＠gmail.com 30 Oct '09

30 Oct '09

Hi, I've tried to use ffplay as a shared library. I don't know if I've done things properly, I've brutally replace main in ffplay.c by some function name, add fPIC to CFLAGS and instead of making a binary of ffplay I've build a shared lib. First question : how good is that ? Then I've linked my application to ffplay.so. My goal is to intercept all the dynamic memory management (malloc, calloc, free, ...). To do that I've preload a library that intercept malloc (and other allocation function). What I'm doing in those interceptors is just to increase a variable to keep trace of the allocated memory. So my problem is that with this technique I observe that only 1.6MB of memory is used (when I use top to monitor the memory used by the ffplay process I got something like 200 MB). Those values are when displaying a locally stored H.264,1080p movie. Could someone explain me what's going on ? Where did I mess up ? thanks a lot, Maxime -- Maxime Louvel +33 6 85 91 05 30 3 rue charreton 38000 Grenoble France

2 2

Sorenson Spark Decode
by ed.emre＠gmail.com 30 Oct '09

30 Oct '09

Hi, Ffmpeg can decode sorenson spark video codec? i dont have any video that encoded with sorenson spark. thanks.

3 5

Multicore use in ffmpeg
by lists＠mrnaz.com 30 Oct '09

30 Oct '09

Hi all, At the moment, when encoding, top reports ~90% cpu use for the ffmpeg process, but ~45% cpu use overall. No other processes are using a notable amount of CPU, the next highest is x using about 2%. If I'm interpretting this right, ffmpeg is only encoding with one of the two cores in the machine. Is that correct ? - Naz.

2 4

Re[Ffmpeg-user] quire sample .avi file and .wmv file
by somesh2010＠gmail.com 30 Oct '09

30 Oct '09

Hi, Can anyone give me a link/site where I can download some legal samples, of .avi file and .wmv file. Thanks usr321 -- View this message in context: http://old.nabble.com/Require-sample-.avi-file-and-.wmv-file-tp26095320p260… Sent from the FFmpeg-users mailing list archive at Nabble.com.

1 0

apt-get install ffmpeg
by yihai2004＠gmail.com 29 Oct '09

29 Oct '09

hello, I use ubuntu 8.10, the simplest way to install ffmpeg is to use "apt-get install ffmpeg", but what's if I want to customize the configuration to ensure ffmpeg support certain codec? like x264? Can I still use "apt-get" command to do the customization? Or the only way is to "./configure make make install " from the source? Thanks a lot! Hai

14 25

Problem of adding a logo to decoded YUV data.
by newptcai＠gmail.com 29 Oct '09

29 Oct '09

Please help! This a code snippet to add a logo to decoded YUV picture (by ffmpeg). http://pastebin.com/md6c16c5 It almost works. Only one problem : sometimes the logo's board is heavily messed up with the surrounding part. The most strange thing is that sometimes it works OK but sometimes it doesn't The code would modify the AVFrame allocated by ffmpeg. Would that cause the problem? //--------- void OsdLogo(AVFrame* picture, unsigned char *pBufLogo, int xPos, int yPos) { unsigned char * pYBuf = picture->data[0]; unsigned char * pUBuf = picture->data[1]; unsigned char * pVBuf = picture->data[2]; unsigned char * pBufLogoY = pBufLogo; unsigned char * pBufLogoU = pBufLogo+180*52; unsigned char * pBufLogoV = pBufLogoU+180*52/4; int i,j; //xPos ??? if ((xPos&0x01)!=0) { xPos&=(~0x01); } if ((yPos&0x01)!=0) { yPos&=(~0x01); } pYBuf+=(yPos*picture->linesize[0]+xPos); pUBuf+=(yPos*picture->linesize[0]/4+xPos/2); pVBuf+=(yPos*picture->linesize[0]/4+xPos/2); for (i=0; i < 52 ; i+=2) { for(j=0; j < 180 ; j+=2) { if ((*pBufLogo)<205) { //Y *(pYBuf+picture->linesize[0]) = *(pBufLogo); *pYBuf++ = *pBufLogo; *(pYBuf+picture->linesize[0]) = *(pBufLogo); *pYBuf++ = *pBufLogo; //U (*pUBuf++) =(*pBufLogoU++); //V (*pVBuf++) =(*pBufLogoV++); } else { pYBuf+=2; pUBuf++; pVBuf++; pBufLogoU++; pBufLogoV++; } pBufLogo+=2; } pYBuf+=(picture->linesize[0]-180+picture->linesize[0]); pUBuf+=((picture->linesize[0]-180)/2); pVBuf+=((picture->linesize[0]-180)/2); pBufLogo+=180; } } -- look to the things around you,the immediate world around you, if you are alive,it will mean something to you ??Paul Strand

1 0

Feed from FLV stream to rtp stream
by Frederic.Paquette＠tatacommunications.com 29 Oct '09

29 Oct '09

Hello, I have the following problem, I need to take as an input an http FLV stream and convert it to an rtp stream. The problem is that there seem to be no way to send the audio track onto one RTP output stream and the video stream onto another RTP stream and you cannot send two stream onto the same rtp port. In fact, I did find a way, it is to start two separate instances of ffmpeg, one grabbing the video and the second grabbing the audio and both pushing to different rtp ports. However, the issue is that both resulting streams are using the same payload type, so when I merge those in the SDP file on the streaming server, the player is unable to play it properly because of the payload type conflict. Both RTP stream use the payload type 96. The codecs I use for the output are H.263 and AMR 10.2K. The is no parameter that would allow the user to change the payload type at the output of ffmpeg on rtp. If someone has the recipe to my problem, i would be interested to know. thanks and regards FFmpeg version git-7291266, Copyright (c) 2000-2009 Fabrice Bellard, et al. built on Oct 27 2009 15:07:43 with gcc 4.1.2 20070925 (Red Hat 4.1.2-27) configuration: --enable-gpl --enable-version3 --enable-pthreads --enable-libopencore-amrnb --enable-libopencore-amrwb --enable-libx264 --enable-libmp3lame --enable-libfaad --enable-libfaac --enable-nonfree --enable-shared libavutil 50. 3. 0 / 50. 3. 0 libavcodec 52.37. 1 / 52.37. 1 libavformat 52.39. 2 / 52.39. 2 libavdevice 52. 2. 0 / 52. 2. 0 libswscale 0. 7. 1 / 0. 7. 1 FFmpeg git-7291266 libavutil 50. 3. 0 / 50. 3. 0 libavcodec 52.37. 1 / 52.37. 1 libavformat 52.39. 2 / 52.39. 2 libavdevice 52. 2. 0 / 52. 2. 0 libswscale 0. 7. 1 / 0. 7. 1 <mailto:e at tatacommunications.com>

1 0