[Ffmpeg-devel-irc] ffmpeg.log.20141201

burek burek021 at gmail.com
Tue Dec 2 02:05:01 CET 2014


[00:16] <MadTBone_> Looking through demuxing_decoding.c, there are several pointers declared that only reference things from the AVFormatContext.  Like lines 256 and 257:   --  video_stream = fmt_ctx->streams[video_stream_idx];  video_dec_ctx = video_stream->codec;   --  then  video_dec_ctx  is used to get the height, width, and pix_fmt.  Other than brevity and code readability, is there any good reason to do this?  (not that those aren't good reasons!)
[00:17] <MadTBone_> That is, rather than getting those values like:   fmt_ctx->streams[video_stream_idx]->codec->width
[00:26] <emhs> Command and output: http://pastebin.com/tysFEBE7
[00:30] <emhs> I'm trying to get this to stick a large, wide image behind a narrow video.
[00:53] <MadTBone_> emhs, hold on a sec
[00:54] <MadTBone_> I do this all the time, but with 2 videos....
[00:54] <emhs> MadTBone_: I've actually just converted the image to a video to make it simpler.
[00:55] <emhs> I've now got backdrop.mp4, which is 3414x1920 and Necessity.mp4 which is 1080x1920.
[00:55] <MadTBone_> you don't have to, and you'll get better quality referencing the original image (unless you used a lossless video codec)
[00:57] <emhs> MadTBone_: Latest attempt. Both this way, and with just the image have been producing nothing but dropped frames, and clean audio. http://pastebin.com/fjFgy4yS
[01:13] <MadTBone_> emhs: ffmpeg -loop 1 -i image.png  -i video.mp4 '[0:v][1:v]overlay=x:y' -shortest out.mp4
[01:14] <t4nk639> if one is to encode a frame of audio, does it have to match the AV_SAMPLE_FMT of the encoder?
[01:14] <emhs> MadTBone_: Trying it now.
[01:15] <emhs> MadTBone_: Doesn't that need the -filter_complex?
[01:15] <MadTBone_> emhs:  you don't need the [0:v][1:v] convention... I usually use it for chaining together multiple filter paths.
[01:15] <MadTBone_> oh... yep
[01:15] <MadTBone_> mistyped
[01:15] <emhs> No worries, thanks.
[01:17] <emhs> Woot!
[01:17] <emhs> MadTBone_: Nailed it!
[01:17] <emhs> Dude, thanks for that.
[01:17] <MadTBone_> no prob.
[01:17] <MadTBone_> it was just the -loop 1
[01:18] <MadTBone_> might have probably been able to do something with setpts as well in the filtergraph
[01:18] <emhs> I _KNEW_ it was some stupid finicky thing I'd missed. Thanks for spotting it, dude.
[01:26] <MadTBone_> t4nk639, yes.  if not, run it through libswresample.
[01:27] <t4nk639> MadTBone_: thanks
[01:27] <t4nk639> does anyone know that the "extended_data" ptr is for audio?
[01:29] <MadTBone_> t4nk639: planar audio channels.   https://www.ffmpeg.org/doxygen/trunk/structAVFrame.html#afca04d808393822625e09b5ba91c6756
[01:29] <MadTBone_> (or packed)
[01:30] <MadTBone_> remember, it's a uint8_t**  (not a uint8_t*)
[01:31] <MadTBone_> so, you can get the pointer to the second data plane (for planar audio) by doing something like   *myframe->extended_data + 1
[01:31] <t4nk639> excellent, thanks
[01:33] <MadTBone_> also, remember that packed audio (most plain PCM formats, like .wav) have all channels interleaved in first data plane
[01:34] <t4nk639> will swr_convert automatically handle converting interleaved samples to planar if i feed it the right parameters?
[01:35] <t4nk639> still trying to resolve a segfault, so not sure it works yet :P
[01:41] <MadTBone_> it should go between packed and planar, assuming everything's inited properly.  what are your in/out fmts?  how are you alloc'ing the frame's buffers? are you (de)muxing or dealing with raw codecs?
[01:42] <t4nk639> s16le to fltp
[01:42] <t4nk639> was the built in aac encoder from a raw pcm stream
[01:45] <t4nk639> av_samples_alloc_array_and samples
[01:45] <t4nk639> was what i was using (according to the example in the doc/)
[01:45] <MadTBone_> take a look at resampling_audio.c ( https://www.ffmpeg.org/doxygen/trunk/resampling__audio_8c.html )  ... it shows how to do  AV_SAMPLE_FMT_DBL  to AV_SAMPLE_FMT_S16, but it should be easy to go to/from any.
[01:46] <t4nk639> seems to work when i use the example, i just need to track down an error in my code
[01:46] <MadTBone_> ok... you already know about the examples in the docs
[01:46] <MadTBone_> valgrind or gdb
[01:47] <MadTBone_> go with gdb first, as valgrind will make you depressed with everything it catches
[01:47] <t4nk639> yeh, will make an x86 version of it and debug it, was trying to do it on android on which debugging is a pain
[01:47] <MadTBone_> (it'll also make your code run many times slower
[01:47] <MadTBone_> yep
[01:48] <MadTBone_> using the emulator, or on the hardware?
[01:48] <t4nk639> hardware
[01:48] <t4nk639> i can actually debug it on the hardware, but it's just a laborious process
[01:48] <t4nk639> and i don't have debug symbols enabled
[01:49] <t4nk639> so will need to rebuild one and see if i can spot my error on linux or something :)
[01:49] <MadTBone_> good luck
[01:49] <t4nk639> thanks
[01:49] <t4nk639> first time playing around with ffmpeg, quite a lot to take in :)
[01:55] <t4nk639> also, if encoding audio, do i also need to match the sample rate of the encoding format?
[01:55] <t4nk639> or is that handled by encode
[02:11] <MadTBone_> t4nk639:  sample rates need to match. again, libswresample will let you match sample rates
[02:13] <t4nk639> thanks again!
[02:14] <t4nk639> just found the transcode_aac.c example, seems very informative, i should of studied that before i wrote this mess :} http://ffmpeg.org/doxygen/trunk/transcode_aac_8c-example.html#a85
[02:14] <t4nk639> but learning as i go, so it's all good.
[02:16] <t4nk639>  well, gonna take a greak thanks MadTBone_  for your help.
[04:13] <k_sze> What's the difference between yuyv422 and yuv422p?
[05:47] <Borys> Hi. Which software I should use on remote vps to broadcast rtp audio over internet from my local pc to few mobile devices on android/ios? I need it for my Fire Departament
[11:21] <marcosf0> Current RPM Fusion repository description (website)  is "<strong>Redhat</strong> and <strong>Fedora</strong> packages for EL, Fedora Rawhide (i386, x86_64)"
[11:22] <marcosf0> I suggest simplify it to "<strong>Fedora</strong> and <strong>Red Hat Enterprise Linux</strong> packages (i386, x86_64)"
[11:24] <relaxed> marcosf0: take that up with the packager
[11:25] <marcosf0> this is on ffmpeg.org download section
[11:25] <relaxed> or the webmaster for that site
[11:26] <relaxed> it's listed for convenience- ffmpeg is not responsible for RPM Fusion builds in any way.
[11:28] <marcosf0> it is the descripton on the ffmpeg.org website: https://github.com/FFmpeg/web/blob/master/src/download#L69
[11:28] <relaxed> oh, haha, sorry
[11:30] <relaxed> marcosf0: create a patch and submit it to the bug tracker as a feature request
[11:42] <marcosf0> ok, thanks
[11:49] <hay> hi, why do I get new stream every second or so (in VLC Player's Codec information) when trying to play muxed live MP3 stream along with the UScreenCapture using the following command: ffmpeg -f dshow -i video="UScreenCapture" -re -i "http://82.149.22.34:8000/CityMp364mono.mp3" -acodec copy -vcodec mpeg2video -s 720x576 -qscale:v 2 -r 10 -f mpegts -b:v 700k -minrate:v 700k -map 0:0 -map 1:0
[11:49] <hay> udp://192.168.0.106:1230?pkt_size=1316?buffer_size=65535
[11:49] <hay> any clarifications or ideas are very welcome after I spent quite some time on this :)
[12:43] <hay> oh, just moved further on... it seems that even when I play a saved .ts file as -f mpegts this way: ffmpeg -i "San_Diego_Clip.ts" -f mpegts udp://192.168.0.106:120, VLC shows new streams constantly adding up - is this normal? When playing .ts file directly in VLC, there are just two streams (one audio and one video)... is this behaviour normal?
[14:48] <Aartsie> hi all
[14:49] <Aartsie> is it true that ffmpeg is changed to avconv ?
[14:50] <sacarasc> avconv/Libav is a fork. ffmpeg is still around.
[14:50] <sacarasc> Read that for more info.
[14:51] <sacarasc> Also, Debian has ffmpeg back in the repos, so maybe Ubuntu will follow suit soon.
[14:52] <Aartsie> oke my raspbian gives that message ;)
[14:52] <Aartsie> wich is debian
[14:52] <sacarasc> See the end of the topic. ;D
[14:52] <Aartsie> yeah i have installed ffmpeg now
[14:53] <Aartsie> the strange thing is that it won't work like my archlinux box
[14:54] <Mavrik> yes
[14:54] <Mavrik> because Debian installas avconv when you install ffmpeg package P
[14:54] <Aartsie> Mavrik: how is that possible ?
[14:55] <Mavrik> Because... politics.
[14:55] <Mavrik> It's back in latest Debian unstable and Ubuntu 15.04
[14:55] <Aartsie> oke but when i add sudo add-apt-repository ppa:jon-severinsson/ffmpeg it will work on ARM ?
[14:57] <sacarasc> Probably not... You might have to just use avconv/Libav until Raspian/Debian has ffmpeg in more repos.
[15:02] <BtbN> Sid already has ffmpeg? Or still just experimental?
[15:04] <relaxed> BtbN: sid
[15:11] <Aartsie> hmmm that s*cks :P now i have to use avconv :S
[15:12] <Aartsie> it have more options i see
[15:19] <Aartsie> ffmpeg will start up but it won't stream over rtmp :(
[19:44] <realtybaron> i'm attempting to build on a macos (mavericks) and it appears build fails with "ERROR: freetype2 not found".  i've installed freetype via homebrew, but build continues to fail. where is freetype2 expected to be found?
[19:58] <anshul_mahe> check config.log
[20:05] <realtybaron> http://pastebin.com/Ykg1TQVt
[20:06] <realtybaron> "error: cannot find -lfreetype" and then "error: undefined reference to 'FT_Init_FreeType'"
[21:15] <jdm07> is the stagefright plugin the fastest way to playback hw accelerated video on android using ffmpeg?
[23:01] <wyatt8740> When encoding a SWF, I get the message "audio fifo too small to mux audio essence" "av_interleaved_write_frame(): Operation not permitted"
[23:01] <wyatt8740> what does that mean?
[23:06] <llogan> wyatt8740: i don't know, but the command and console output would be helpful
[23:07] <wyatt8740> ffmpeg -i 01\ Conquistador\ \(Mono\).mp3 -i 01\ Conquistador\ \(Mono\).mp3 -map 0:0 -map 0:1 -acodec copy Conquistador.swf
[23:07] <wyatt8740> this mp3 has a album cover on it
[23:08] <wyatt8740> http://pastebin.com/vZW2Fysg
[23:08] <wyatt8740> also tried with just the one -i and without the mapping
[23:08] <wyatt8740> same result
[23:09] <wyatt8740> (the song is "Conquistador" by Procol Harum. Copyrighted so I can't really upload it legally)
[00:00] --- Tue Dec  2 2014


More information about the Ffmpeg-devel-irc mailing list