[Ffmpeg-devel-irc] ffmpeg.log.20130626

[burek]

[00:06] <superware> I'm getting "Option mpegts_original_network_id not found." yet documentation says it exists
[00:20] <llogan> superware: your ffmpeg/ffplay may be too old for the docs shown on the web site. refer to your local docs or upgrade.
[00:24] <superware> llogan: "ffplay version N-54178-gbbe26ef ... built on Jun 24 2013"
[00:27] <superware> llogan: http://pastebin.com/DDZdj08u
[00:29] <superware> no visible video. yet vlc.exe udp://@:1234 works.
[00:30] <llogan> why did you mention mpegts_original_network_id but then paste something else?
[00:30] <superware> my guess is probing is poor in this case, so I need to somehow hint
[00:31] <superware> no problem, a sec
[00:32] <superware> here goes http://pastebin.com/tSYLj7dM
[00:33] <llogan> try increasing buffer_size: http://ffmpeg.org/ffmpeg-protocols.html#udp
[00:34] <superware> ... udp://0.0.0.0:1234?buffer_size=100000 - same
[00:39] <superware> llogan: any idea? :|
[00:39] <llogan> no
[00:45] <superware> :(
[03:21] <AisIceEyes> Hi! I'm a relatively new user of ffmpeg commandline and I'm trying to normalize an audio from a video. I saw this thread through searching - http://ffmpeg.org/pipermail/ffmpeg-devel/2013-March/140826.html - And I'm wondering where do I get the script or where was it pushed to? Thanks!
[03:28] <pandeiro> i'm trying to add srt subtitles to an avi file with `ffmpeg -i original.avi -vf subtitles=movie.srt out.avi` and it works as expected, but the output file is about 1/10 the size/quality. what do i need to change?
[03:36] <AisIceEyes> pandeiro: you have to use an codec I think. Since it's an avi container, maybe XviD?
[03:36] <AisIceEyes> pandeiro: http://ffmpeg.org/ffmpeg.html#Video-Options
[03:37] <klaxa> pandeiro: ffmpeg encodes by default, to keep the original tracks, add: -c copy
[03:37] <klaxa> oh wait...
[03:37] <klaxa> you are hardsubbing, never mind
[03:46] <pandeiro> yeah i am hardsubbing so i can watch an avi with subs on a tv without srt support
[03:47] <pandeiro> AisIceEyes: is there no way to just use the same encoding as the original avi?
[03:49] <klaxa> you can check what codec was used and use similar values
[03:50] <klaxa> you can copy the audio at least
[03:50] <klaxa> so there will be no quality loss
[03:51] <AisIceEyes> pandeiro: What klaxa said :)
[03:51] <pandeiro> ok
[03:51] <pandeiro> any chance you could walk me through that this time? :)
[03:52] <klaxa> pastebin what ffprobe -i original.avi outputs
[03:53] <AisIceEyes> Hmm.... I've been wondering, is there a normalize filter in audio filters in ffmpeg? Maybe "normalize" is not the term?
[03:53] <bencc> I'm using the following command:
[03:54] <bencc> ./ffmpeg -i rtmp://127.0.0.1/audio/test -ar 44100 -f mp3 pipe:1
[03:54] <bencc> I only see the output when stopping the rtmp stream
[03:54] <bencc> how can I make ffmpeg output result in real time?
[03:54] <klaxa> AisIceEyes: see http://ffmpeg.org/ffmpeg-filters.html for all filters, maybe you'll find something?
[03:55] <bencc> klaxa: hi
[03:55] <bencc> klaxa: any idea how to use the rtmp_live option from here? https://lists.ffmpeg.org/pipermail/ffmpeg-cvslog/2012-May/050143.html
[03:55] <klaxa> bencc: hi, no idea, sorry :(
[03:55] <bencc> just the syntax
[03:56] <pandeiro> klaxa: http://sprunge.us/XbZZ
[03:56] <klaxa> hmm still no idea, i've never dug through the ffmpeg/libav* source
[03:57] <klaxa> pandeiro: try this: ffmpeg -i original.avi -vf subtitles=movie.srt -c:v libxvid -b:v 1900k -c:a copy out.avi
[03:58] <pandeiro> klaxa: thanks, will do. curious: can i read about those on ffmpeg's man page or would they be described somewhere else?
[03:58] <klaxa> read about what? the parameters? they should be documented in the manpage
[03:59] <klaxa> in general the documentation on the website is maybe even more elaborate
[03:59] <klaxa> http://ffmpeg.org/ffmpeg.html
[03:59] <AisIceEyes> klaxa: Actually, I did checked that before going to this channel but to no avail. There is however a python script I found in the mailing list - http://ffmpeg.org/pipermail/ffmpeg-devel/2013-March/140826.html - which I am not sure how to use or check Hmm...
[04:00] <klaxa> i don't know how normalization works at all, maybe you could outsource the normalization to sox
[04:00] <sacarasc> AisIceEyes: The script is found in the tools section of the source download.
[04:00] <pandeiro> klaxa: ok. command is running but i notice some output: Neither PlayResX nor PlayResY defined. Assuming 384x288
[04:00] <pandeiro> is that relevant?
[04:00] <klaxa> mmh probably?
[04:01] <pandeiro> yeah seems :)
[04:01] <klaxa> abort now, add -t 30
[04:01] <klaxa> then try to play the output
[04:01] <pandeiro> add -t 30 to the command and run it?
[04:01] <klaxa> yes
[04:02] <klaxa> that will encode only the first 30 seconds so you can check whether or not it will work properly
[04:02] <pandeiro> ah gotcha
[04:04] <pandeiro> seems like it worked
[04:04] <klaxa> ah good, then remove the -t 30 and encode the complete video
[04:05] <pandeiro> that same ffprobe command should show the same resolution in the Stream #0.0 line?
[04:05] <pandeiro> 720x400 in this case
[04:05] <klaxa> then it should all be good
[04:05] <pandeiro> cool thanks very much
[04:05] <klaxa> np :)
[04:23] <RigidWig> Say I wanted to have ffmpeg split and cut out segments of a video based on how dark (how much black, I suppose) those image frames contain (the video will be assmebled from jpegs), would ffmpeg have a native capacity to do this, or would I have to do that elsewhere and just serve up the array of selected files to ffmpeg?
[04:26] <highgod> Hi, I want to ask a quesion.how can I get the number of slices in on frame? slice_group_count? thanks
[04:40] <t4nk283> Hello there is a serious bug in ffmpeg that has been there for ages and i'm not having any luck getting anyone to look at it - if anyone is listening, interrupt_callback is not being fired at all anymore by any of the blocking methods which means that if you try to connect to a broken stream in windows and you dont specify a timeout then ffplay will hang indefinitely.
[08:46] <t4nk327> ffmpeg is no longer calling interrupt_callback on blocking functions - any ideas? I've tried reporting it as a bug but no-one seems interested....?
[08:59] <adnap> Hi
[09:00] <adnap> How would I stream a video in .mkv format from one linux computer to another and play it using mplayer2?
[09:00] <adnap> Also, is it possible to seek while streaming?
[09:20] <superware> should ffplay experience difficulties with an h.264-ts-udp stream?
[09:41] <superware> can someone please help me? http://pastebin.com/ZyNxavHJ
[10:22] <superware> Mavrik: hi
[10:34] <superware> can someone please help me with an h264-ts-udp stream (please see  http://pastebin.com/ZyNxavHJ), it's perfectly playable using VLC... :|
[10:36] <Mavrik> [udp @ 0180c820] Part of datagram lost due to insufficient buffer size
[10:36] <Mavrik> this is your problem it seems.
[10:52] <superware> Mavrik: I've also tried udp://0.0.0.0:1234?buffer_size=100000 but it's the same
[10:53] <Mavrik> 100KB buffer?
[10:53] <superware> I tried other sizes, it doesn't matter
[10:54] <superware> ffplay can play the stream after I save it (raw) to a .ts file using VLC
[10:55] <superware> the transmitter is doing something above the TS stream (udp packets etc) that VLC can handle, but ffmpeg not
[10:55] <superware> maybe..
[10:55] <Mavrik> no
[10:55] <Mavrik> the error is very clear
[10:56] <Mavrik> your UDP buffer is being overrun and your packets are being corrupted
[10:56] <Mavrik> VLC just has a bigger buffer it seems
[10:56] <superware> how should I make the buffer larger?
[10:58] <superware> http://ffmpeg.org/ffplay-all.html#udp
[11:02] <superware> it seems fifo_size is very much related, now the errors are mostly "non-existing PPS 0 referenced"
[11:19] <anshul_>  i have some memory leak, when i am using livavformat.   i do have valgrind lo, but i dont knoe where to paste
[11:22] <anshul_> i have pasted the valgrind log over http://pastebin.com/ZAE83eaR
[11:25] <microchip_> anshul_: if you think it's a genuine memory leak, open a bug report on ffmpeg's trac
[11:32] <anshul_> how to report bug, ffmpeg's trac ???
[11:33] <saste_> anshul_, ^^
[11:35] <anshul_> got it  !
[11:41] <dddh> is it expected to have cod running on GPU in near future?
[11:41] <dddh> *code
[11:56] <superware> Mavrik: can I unicast udp over the internet?
[11:56] <Mavrik> yep
[11:56] <Mavrik> dddh, code running on GPU?
[11:59] <superware> Mavrik: can I stream is to you? :) I'll make it small bitrate
[11:59] <Mavrik> er, why?
[12:00] <superware> Mavrik: see if you're able to see it
[12:04] <dddh> Mavrik: yes
[12:05] <Mavrik> dddh, can you explain more about what do you mean about that?
[12:15] <dddh> Mavrik: cuda computing?
[12:16] <Mavrik> ah, no, there are no bigger plans, CUDA isn't really good for quality video encoding.
[12:16] <Mavrik> there are some plans to integrate platform native libraries, but aren't very active
[12:17] <TheSchaf> cuda sucks for stuff where you have to use ram
[12:17] <TheSchaf> because cpu ram <=> gpu ram is the worst bottleneck
[12:18] <Mavrik> and since execution units are limited on which ram they can access
[12:18] <Mavrik> which really sucks for video-type pattern search on large frames :)
[12:19] <dddh> heh
[12:20] <dddh> current GPUs have enough RAM and there is dynamic parallelism since compute capability 3.5
[12:21] <dddh> you could load more data from host to device memory
[12:22] <JEEB> there's a reason why everyone has started putting hardware encoders on GPUs instead of trying to use more GPGPU
[12:23] <JEEB> MultiCoreWare together with AMD's support did do an OpenCL lookahead ME for x264, but 1) it is lower quality than the CPU algorithm, and 2) it only gives a ~10% increase of speed
[12:23] <JEEB> GPGPU is good for some stuff, but not for the most used formats' video encoding
[12:24] <dddh> I remember that story
[12:25] <dddh> but my GPUs are more expensive than my CPU
[12:25] <JEEB> that still doesn't change the fact that they are good for some stuff, and worse for other
[12:26] <dddh> JEEB :)
[12:27] <JEEB> and it's already a really good result that the lookahead ME is actually making it even a bit faster, even if it is of lower quality
[12:27] <JEEB> anyways, I recommend putting GPGPU to use where it can actually be more useful :)
[12:27] <JEEB> there are various things those things are good at
[12:27] <Mavrik> I'd rather use efforts to integrate with platform encoding libraries
[12:27] <Mavrik> way more useful and faster
[12:28] Action: Mavrik puts "fix libstagefright" on TODO. :P
[12:28] <JEEB> yeah
[12:28] <JEEB> proper HW enc/dec support for the SOCs on mobile devices is a good thing
[14:48] <FuseOnFire> ffmpeg-devel
[14:54] <FuseOnFire> regarding ffmpeg for android - can that be used to stream/play rtsp streams (like with ffplay)?
[14:56] <Mavrik> yes, it's not easy though.
[14:56] <Mavrik> performacne isn't stellar either
[14:58] <FuseOnFire> essentially I'm looking for a way to play rtsp streams with a low-delay (the android built-in videoView has latency in seconds in LAN)
[14:58] <Mavrik> hmm
[14:58] <Mavrik> what's your android target platform_
[14:58] <Mavrik> ?
[14:59] <FuseOnFire> 4.x I guess
[14:59] <Mavrik> your problem is basically severalfold:
[14:59] <Mavrik> 1.) Alot of CPU's aren't fast enough for decoding and burn ALOT of power, ffmpeg doesn't have a working HW decoder support on Android
[15:00] <Mavrik> 2.) Once you decode video you need to get frames on screen which can be pretty slow
[15:00] <zap0> android doesn't have hardware.
[15:00] <JEEBsv_> I think with 4.1/2 you could finally only use the HW decoder, so he could use ffmpeg or whatever to grab the data / demux, and then push that data to the HW decoder in his app
[15:02] <Mavrik> yeah, that's why I asked about android platform version
[15:02] <FuseOnFire> would ffmpeg be able to handle the rtsp streaming as ffplay can?
[15:02] <Mavrik> since 4.1+ has APIs for HW decoder&. BUT.
[15:02] <Mavrik> those APIs are mostly useless for widespread usage ATM :)
[15:02] <Justin> Anyone have familiarity with using ffmpeg and direct show input?
[15:04] <zap0> DS is shite.    most hardware vendors provide their own libs.
[15:04] <Justin> I know. I'm having the strangest issue.
[15:05] <Mavrik> if anyone figures out how to detect which frame format is expected by the en/decoder on Android 4.1, I'd be very grateful though ;)
[15:05] <zap0> the fact you think it should actually work is strange.
[15:05] <Mavrik> zap0, are you just here for random trolling?
[15:05] <zap0> no.
[15:05] <Mavrik> Justin, it's easier if you just tell us what the issue is :)
[15:06] <zap0> i have years of experiance with DS.  i know its shite.
[15:07] <Justin> Direct show directly to rtp multicast stream results a lot of dropped packets. Audio log shows they're old. If I take direct show to file, plays perfectly. I can then stream the file without a problem. If I take the direct show input to rtp, grab that stream with ffmpeg again and restream without re-encoding, the network speaker plays perfectly.
[15:07] <Mavrik> hmm
[15:07] <Mavrik> Justin, that looks more like network buffering issue than anything connected with DShow
[15:09] <Justin> Here's the command line that I'm using. I've tried several other switches. Maybe you have a suggestion? ffmpeg -loglevel debug -f dshow -c:a pcm_s16le -ac 1 -ar 22050 -vn -i audio="Line In (USB Multi-Channel Audi" -c:a copy -f rtp "rtp://239.20.20.6:20222"
[15:17] <Justin> Perhaps there's a buffer switch I've over looked that could be used?
[15:18] <Mavrik> hmm
[15:18] <Mavrik> I wouldn't know of any on RTP
[15:19] <Justin> max delay is ignored by of framesize = 0, bufsize seemingly has no effect. async helps a little
[15:20] <Justin> and the fact that if I grab the stream and rebroadcast, it just works just makes me a little insane.
[15:21] <Justin> It's also not codec dependant. i encoded to ulaw and it does the same.
[15:31] <nwave> is it possible to serve an rtmp stream from ffmpeg
[15:50] <Mavrik> nwave, no, you'll need a streaming server.
[15:51] <nwave> ok, I'm playing with crtmpserver.  Having a bit of trouble doing a simple test on it.  I assume I can just push an rtmp feed from ffmpeg to the streaming server?
[15:54] <Mavrik> yep, that you can do
[15:55] <nwave> http://pastebin.com/tuWjLUpt
[15:56] <nwave> this is the crtmpserver error log when I try to send data to it
[16:08] <Nunix> Hi all. I'm trying to decode an H264/MP3 RTMP live stream but I can't get my audio and video to sync. I can see after av_read_frame() that, after decoding, the audio and video PTS is sequential and appears correct but the actual audio that I hear is about 10 secs late. The video is perfect. Almost instantaneous. Is there any audio delay or buffer that I need to switch off?
[17:10] <nwave> is it possible to publish to an rtmp server as well as saving the stream to disk using the segmenter
[17:12] <JEEBsv_> you can have multiple outputs
[17:13] <JEEBsv_> just set the first output file name/url/whatever and then start setting output setting again and then a second output file name/whatever
[17:40] <nwave> this seems like it should work, but it's giving me http://pastebin.com/jNFg4jq5
[17:42] <JEEBsv_> well it is having a problem with the first output you are setting
[17:42] <JEEBsv_> hmm
[17:43] <JEEBsv_> is that segs/stream%05d.ts way of setting the output file name correct?
[17:45] <nwave> yeah it works fine when used without the second output setting
[17:45] <nwave> only got this error after trying to combine the outputs
[18:50] <nwave> does anyone understand how the segment_list is created
[18:51] <nwave> when I have segment_wrap enabled it just makes a really long file and never gets rid of old file entries
[19:58] <t4nk096> Hi I have written an application using libavcodec (more specifically libx264) and am facing a segmentation fault issue that I am unable to debug. Would this be a correct forum to to enquire about the same?
[20:20] <whateverman> so, question
[20:20] <sdl240> 42!
[20:20] <whateverman> clever
[20:20] <whateverman> I was still typing
[20:20] <whateverman> I've got some videos I'm concatenating, and I'm running into issues with the second video's audio dropping out
[20:21] <whateverman> I'm converting the h264 encoded videos to ts files then using the concat filter
[20:21] <whateverman> the second video has a mono audio track - how do I have it just use that as the L+R track and an empty L-R track for the latter part of the video?
[20:52] <whateverman> so, is there a sane way to take a mono audio channel and upmix it to stereo for the concatenation
[21:01] <llogan> whateverman: i may have missed an earlier discussion, but you can take a mono input and turn it into a stereo output with -map_channel or the pan audio filter.
[21:01] <whateverman> ah
[21:02] <whateverman> maybe you could help me with what I'm doing here
[21:02] <whateverman> I'll get my scripts all together to show you if you'd care to take a look
[21:02] <llogan> ok
[21:03] <whateverman> http://hastebin.com/haduloqibe.bash
[21:04] <whateverman> ffmpeg info
[21:04] <whateverman> http://hastebin.com/lapukokeye.vhdl
[21:04] <whateverman> this works to give me a video file of proper length without audio
[21:05] <llogan> i'll just give you some examples
[21:05] <llogan> ffmpeg -i mono.wav -map_channel 0.0.0:0.0.0 -map_channel 0.0.0:0.0.1 stereo.wav
[21:05] <llogan> or
[21:05] <llogan> ffmpeg -i mono.wav -af pan=2:c0=c0:c1=c0 stereo.wav
[21:05] <whateverman> ah, ok
[21:06] <mark4o> won't -ac 2 work if you just want to copy it to both channels?
[21:07] <llogan> heh, yes.
[21:07] <llogan> but what fun is that.
[21:09] <llogan> whateverman: did you also see the concat demuxer? http://ffmpeg.org/ffmpeg-formats.html#concat-1
[21:09] <llogan> https://ffmpeg.org/trac/ffmpeg/wiki/How%20to%20concatenate%20%28join%2C%20merge%29%20media%20files
[21:09] <whateverman> I did - trying to use that, things took a little long
[21:10] <whateverman> I at least got it downmixing the stereo to mono when concatenating, now
[21:10] Action: llogan curses the lack of CamelCaseUrl
[21:10] <llogan> stereo to mono? i thought you wanted the opposite.
[21:10] <whateverman> (finished version) http://hastebin.com/hibuwikage.bash
[21:10] <whateverman> well, I was thinking that
[21:11] <whateverman> then I realized that this was mainly for phones and tablets
[21:11] <whateverman> I just needed it to match the audio tracks, and I was thinking that data should be preserved
[21:11] <whateverman> but honestly, nobody cares about the audio quality on our videos
[21:11] <whateverman> and phones have one speaker
[21:12] <llogan> do you have any users that use headphones?
[21:13] <llogan> although they probably won't notice either
[21:26] <whateverman> actually, I'm being told that my video crashes browsers? http://alan.appredeemdev.com/out.mp4
[21:26] <llogan> mine just wants to download that
[21:27] <whateverman> odd... works in android
[21:28] <llogan> oh, you meant on a phone. duh.
[21:28] <LithosLaptop> My browser just says the file is corrupt
[21:30] <llogan> worked fine on iphone 3gs.
[21:33] <AisIceEyes> <sacarasc> AisIceEyes: The script is found in the tools section of the source download. --> Would it be alright to ask where exactly? I don't think I see them anywhere here - https://ffmpeg.org/download.html
[21:35] <llogan> download the ffmpeg source then look in the "tools" directory.
[21:38] <AisIceEyes> llogan: Oh! There it is! Thanks!
[21:38] <AisIceEyes> Though would it be alright to ask how to use in ffmpeg line?
[21:39] <llogan> how to use what?
[21:39] <AisIceEyes> *how to use it?
[21:39] <AisIceEyes> I mean the py script
[21:39] <AisIceEyes> Okay, you see llogan, I'm trying to normalize the audio of various files
[21:39] <AisIceEyes> While encoding etc
[21:40] <AisIceEyes> I have first found this through googling - http://ffmpeg.org/pipermail/ffmpeg-devel/2013-March/140826.html
[21:40] <AisIceEyes> Then there is the script
[21:40] <AisIceEyes> I am not sure how to use the normalize.py you see ^^;;
[21:47] <llogan> ./normalize.py input -c:v copy -c:a <your audio encoder> output
[21:47] <AisIceEyes> So it's two commands?
[21:48] <llogan> my example is one command
[21:49] <whateverman> figured out what was up with that file
[21:49] <AisIceEyes> I'm thinking if it's alright if it's possible to be something like this - ffmpeg -i input -filter:v "<can i insert the py here?>" -vcodec something <can i insert the py here?> -acodec something output
[21:49] <whateverman> the audio track was mp2, so I had the last stage explicitly convert audio to aac
[21:50] <AisIceEyes> Or at least call the python script
[21:50] <AisIceEyes> Is it possible by chance?
[21:50] <whateverman> you can always run ffmpeg from python
[21:51] <AisIceEyes> Ah
[21:52] <AisIceEyes> Thanks :)
[21:53] <llogan> AisIceEyes: if you want to use the normalize tool you can use it like it is ffmpeg. just add your options as usual: normalize.py input <output options> output
[21:53] <tkb> hi
[21:54] <llogan> or you can look inside normalize.py and adapt it to your needs
[21:56] <AisIceEyes> llogan: Ah, I think I can do that. Though I don't understand most of the algorithm yet. Will just study it. Thanks :)
[21:57] <AisIceEyes> Just to be clear, there is no way to run the py script in ffmpeg?
[21:57] <llogan> no
[21:57] <AisIceEyes> llogan: Gotcha. Thanks :)
[21:58] <tkb> i got a problem:  when i try to encode an flv video to mp3 with ffmpeg -i vid.flv -acodec libmp3lame -ab 320k out.mp3 (CBR)  it encodes the video to mp3 ...but in some players like audacious or winamp, die duration of the song will be reported wrong and seeking is impossible ... i did read a little bit and found out that the reason is that ffmpeg does not set the xing header in CBR mp3 Files correctly....  how can i set them correctly throu
[21:58] <tkb> gh ffmpeg ?
[22:05] <hypfer> yäy
[22:05] <hypfer> hello everyone
[22:06] <hypfer> i have some broken flac files. mplayer doesn't care, but rhythmbox won't play them. I just want to reencode them with no quality loss.
[22:06] <hypfer> are there special parameters for that?
[22:18] <llogan> hypfer: ffmpeg -i input.flac output.flac
[22:18] <hypfer> hm
[22:18] <hypfer> simple
[22:42] <pyBlob> Is there a way to convert color-mjpgs to grayscale-mjpgs by just leaving out/overriding the U/V-components?
[22:44] <pyBlob> -> no encoding error for Y-component
[00:00] --- Thu Jun 27 2013