[Ffmpeg-devel-irc] ffmpeg.log.20160106

burek burek021 at gmail.com
Thu Jan 7 02:05:03 CET 2016

[00:21:38 CET] <drooly> i have some flac files to transcode to ogg. it seems their enbedded thumbnails are encoded as a videostream with an insanely high framerate.
[00:22:30 CET] <drooly> in any case, when i transcode to ogg (libvorbis), i cannot seek (fast forward/backward) anymore.
[00:22:55 CET] <drooly> discarding video with -vn helps, but i would like to salvage the thumbnail/coverart images.
[00:23:47 CET] <drooly> i am not quite sure if these images are supposed to be an actual videostream, or somehow part of the metadata.
[00:24:12 CET] <drooly> the latter, i tried with https://stackoverflow.com/a/18718265 but failed.
[00:26:12 CET] <drooly> i can salvage one image from the stream, but how can i put it back into the transcoded ogg? i'd also like to be able to script it since i have quite many flacs
[00:29:33 CET] <TD-Linux> maybe try oggenc? see if it does the right thing
[00:30:18 CET] <drooly> TD-Linux: yes, i stumbled over that. trying it now.
[00:30:20 CET] <TD-Linux> normally coverart in ogg is stored in the comment metadata
[00:45:53 CET] <drooly> TD-Linux: oggenc seems to discard the video, and i cannot add the image back in.
[00:46:16 CET] <TD-Linux> drooly, it doesn't keep the album art? it should. what are you using to see the album art?
[00:47:02 CET] <TD-Linux> hmm maybe not. I know opusenc does
[00:53:42 CET] <drooly> i'm looking/listening at the files with vlc and mpv, the result is always the same: the flac works, the ogg either doesn't display coverart and works, or displays coverart and is unseekable.
[00:54:02 CET] <c_14> drooly: you can make it seekable, but it would still be a video stream
[00:54:16 CET] <c_14> ffmpeg doesn't support writing image metadata (at least converting an image to image metadata)
[00:54:40 CET] <c_14> you can do it manually by taking the image and I think base64 encoding it and then setting it as a certain metadata tag, but I forget how
[00:54:43 CET] <c_14> plus it's a pain
[00:55:14 CET] <c_14> If you're fine with it being a separate video stream add something like -vf fps=25
[00:55:17 CET] <drooly> i actually might have to retract my previous theory about the image being a video stream; at least in vlc it displays just fine as metadata.
[00:55:32 CET] <drooly> and no video stream.
[00:55:42 CET] <TD-Linux> if you application can handle opus rather than vorbis, opusenc does handle it correctly
[00:55:57 CET] <drooly> in vlc's handy messages window
[00:56:12 CET] <c_14> drooly: but it's still not seekable?
[00:56:40 CET] <TD-Linux> drooly, is ffmpeg encoding it as a theora video track or something?
[00:56:54 CET] <drooly> the original flac is always seekable, the ogg is only seekable when i see no folder art.
[00:57:21 CET] <c_14> Can you check the output of the conversion that produces the one with folder art?
[00:57:21 CET] <drooly> and no, it's audio only. i think my previous theory was wrong.
[00:57:35 CET] <c_14> If it has something like (0:v -> theora)
[01:08:48 CET] <drooly> i just put together a file with some info, will post a link in a second...
[01:10:19 CET] <drooly>
[01:10:50 CET] <drooly> so, first the command i used for transcoding.
[01:11:01 CET] <drooly> then some info about the original flac.
[01:11:13 CET] <c_14> yeah, it's encoding it as a separate theora video stream
[01:11:15 CET] <drooly> then some info about the resulting ogg
[01:13:03 CET] <drooly> btw, result is the same if i remove "-c:v theora" from the command
[01:14:02 CET] <c_14> yeah, ffmpeg treats image metadata as a separate image stream in the input and then encodes it as video output
[01:15:45 CET] <drooly> hmm. if you look further down, both the original flac and the resulting ogg have a seperate video stream - so this actually isn't the same as metadata?
[01:15:55 CET] <drooly> or is it because ffplay does the same?
[01:16:06 CET] <drooly> when playing the files?
[01:16:35 CET] <drooly> decodes the albumart metadata to a video stream?
[01:16:50 CET] <c_14> yes
[01:18:49 CET] <drooly> basically this means that all my music has been encoded with ffmpeg, which transformed metadata thumbnails into a video stream...
[01:24:27 CET] <drooly> why can't i just tell ffmplay to not create these video streams, and leave the metadata as it is?
[01:27:28 CET] <drooly> and, more importantly, why does it make my file unseekable??
[01:34:04 CET] <c_14> Because there's only 1 video frame and most players have trouble seeking then
[01:35:41 CET] <TD-Linux> if anyone feels like writing code, opusenc's album art stuff could be ported to oggenc. Or there are probably other tools which do it correctly
[01:37:29 CET] <drooly> ah yes, opusenc. what file extension does opus audio have usually?
[01:38:35 CET] <furq> .opus
[01:38:49 CET] <furq> it's an ogg container though
[01:47:55 CET] <drooly> thanks for your help. i guess i'll just extract the images to external files. good night.
[06:03:12 CET] <prelude2004c> Hey guys. . Can someone point me in the right direction. . Stb company tells me that when they record pvr stuff from content I am encoding the ext-e-pvrstarttime  is less than zero. . How did I do that?
[07:55:15 CET] <kichigaimentat> Hello, I've got these H.264 streams encoded by a bunch of knuckleheads that I have no intention of ever interacting with ever again who encoded some stuff at H.264/High Profile/Level5.1 and yet the actual content probably better fits Main/3.1. Any way to redefine the header, even as something like High/4.0?
[07:55:47 CET] <kichigaimentat> As I understand it the Profile (High) defines encoding optimizations, Level defines the bitrates, and the Level is *way* overkill in this case.
[08:05:43 CET] <waressearcher2> kichigaimentat: und bis bald
[10:10:23 CET] <mitch31> hello. please anyone can help me! I use windows to stream my webcam. I stream video to server live.webm but chrome is loading infinite and it is no video. stat says WAIT_FEED
[10:13:53 CET] <k_sze> Does ffmpeg's ffv1 encoder default to level 1 if I don't specify -level?
[10:17:58 CET] <durandal_1707> yes
[10:20:51 CET] <k_sze> I see. So level 3 can be sliced and encoded in a multithreaded way. But how about decoding? Does the level affect whether it can be decoded in a sliced/multithreaded way?
[10:45:27 CET] <k_sze> And what are the pros and cons of using a large context vs a small context for ffv1?
[10:45:37 CET] <k_sze> (the -context argument)
[11:38:36 CET] <matham> I was wondering whether there's a way to set the loglevel specific to a particular AVFormatContext? and similarly codec contexts
[11:39:11 CET] <matham> They all start with a AVClass, so I'm thinking this should be possible?
[11:39:44 CET] <matham> but I don't see any clear api for this. And I couldn't tell whether the AVClass class is shared or specific to the context
[13:29:17 CET] <volar> hi all
[14:34:06 CET] <kynlem> hey
[14:34:51 CET] <kynlem> how come "ffmpeg -i ~/Desktop/VideoForCutting.mp4 -ss 0 -t 60 -vcodec copy -acodec copy ~/Desktop/newchunk00.mp4" produces a video of 1799 frames?
[14:35:22 CET] <kynlem> isn't the first frame always and I-frame? why would it stop 1 frame short (the video is 30 fps)?
[14:35:30 CET] <kynlem> always an*
[14:37:24 CET] <dibalaj> Hello, can anyone help me? I have some issues when restreaming RTSP to YouTube (RTMP).
[14:40:08 CET] <dibalaj> Here is link to ffmpeg forum, if you can please look at it, many thanks! http://ffmpeg.gusari.org/viewtopic.php?f=11&t=2569
[14:56:34 CET] <andy`> kynlem: 60*30/1.001=1798.2
[15:02:29 CET] <Betablocker> dibalaj:  take a look at my query please
[15:02:42 CET] <Betablocker> i love this shit :D
[15:22:43 CET] <termos> is there a way to specify the -re flag when using libFFmpeg?
[15:24:33 CET] <kynlem> Betablocker: where does 1.001 come from?
[15:24:34 CET] <JEEB> there's no libffmpeg
[15:24:48 CET] <JEEB> and there probably is an avoption for it :P
[15:24:55 CET] <JEEB> time to grep what the re option sets :P
[15:26:14 CET] <kynlem> andy`: where does 1.001 come from?
[15:27:11 CET] <termos> libav? seems like it's not part of the library, only part of ffmpeg.c
[15:27:16 CET] <kynlem> and is there a way to just tell ffmpeg to cut 1800 frames, or do i *have* to specify it in seconds?
[15:28:19 CET] <kynlem> not a big deal, but would just feel better to have a holy number of frames rather than unorthodox number of seconds (ocd, i'm looking at you)
[15:28:35 CET] <furq> kynlem: ntsc video is usually 30/1.001 fps
[15:28:52 CET] <furq> or 29.97 if you prefer
[15:30:52 CET] <kynlem> k, i'll try to split video into 60.06-second chunks and see how it goes
[15:30:56 CET] <kynlem> thanks, guys
[15:33:08 CET] <dibalaj> @Betablocker what query do you mean?
[15:34:30 CET] <Betablocker> ;)
[15:57:07 CET] <prelude2004c2> hey everyone good day.. having an issue with playback... ( see http://pastebin.com/XAVQL9bz ) ... basically when i record something from the live stream and i try to playback it always plays the end of the stream and not from the beginning. I am told it is because the start time of the recording has  negative value and the system can't play something with a negative value.. can anyone point me in the right direction
[19:09:47 CET] <prelude2004c2> hey everyone good day.. having an issue with playback... ( see http://pastebin.com/NTF2pjKC ) ... basically when i record something from the live stream and i try to playback it always plays the end of the stream and not from the beginning. I am told it is because the start time of the recording has  negative value and the system can't play something with a negative value.. can anyone point me in the right direction
[19:27:33 CET] <DHE> prelude2004c: playback from the  beginning will happen if the stream finishes and the X-EXT-ENDLIST (exact name?) label is added
[19:27:53 CET] <DHE> also with -hls_list_size 5 you only have the last ~30 seconds of content available anyway
[21:29:20 CET] <lumidee> hi if i want to record a video of my computer screen for a 800x600 VGA beamer, should i change the resolution from 1920x1080p to 1024x768 to have the 4:3 ratio or does it not matter when my LCD is 16:9 anyways
[21:29:52 CET] <lumidee> also iam not sure if a higher resolution will also mean higher quality though i play it at a 800x600 beamer
[21:32:12 CET] <waressearcher2> lumidee: hallo und herzlich willkommen
[21:32:27 CET] <lumidee> waressearcher2, servus
[21:38:17 CET] <DHE> that would make the most sense, changing resolutions. otherwise you will have a distorted image
[21:38:32 CET] <DHE> or would need to crop
[22:19:16 CET] <jafa> seeing an odd problem... invoking avcodec_decode_video2() to decode mpeg2 - I get out a sequence of frames with increasing pts values however the frames seem out of order
[22:19:40 CET] <jafa> if I feed it through a filtergraph it works - frames are in order
[22:20:27 CET] <jafa> pts going into avcodec_decode_video2 jump around following the encoded frame order as expected
[22:21:04 CET] <jafa> pts coming out of avcodec_decode_video2 is in order but the frames don't seem to be
[22:23:23 CET] <Mavrik> jafa, uhm.
[22:23:32 CET] <Mavrik> You should probably feed frames in DTS not PTS order to decoder.
[22:25:18 CET] <jafa> feeding them in natural order from the source
[22:25:47 CET] <Mavrik> "natural order"?
[22:25:58 CET] <jafa> no reordering of data going into the codec
[22:26:20 CET] <jafa> so should already be dts order
[22:27:09 CET] <jafa> ie all jumbled up as per normal mpeg2
[22:27:43 CET] <jafa> works great if I use a filterfraph
[22:28:14 CET] <Mavrik> hrmf, no idea, it should JustWork(tm)
[22:28:22 CET] <Mavrik> Maybe you're looking at a wrong PTS field?
[22:28:42 CET] <Mavrik> there's the av_guess_pts call or whatsit that properly sets PTS on a frame.
[22:29:36 CET] <jafa> using av_frame_get_best_effort_timestamp()
[22:30:07 CET] <jafa> I could use a filtergraph set to "copy" (this works)... not sure what the cost is
[22:30:17 CET] <Mavrik> There has to be some other issue then, worked just fine for my code at least in 2.0 - 2.5
[22:45:49 CET] <jafa> when a filtergraph is set to "copy" does it handle it by reference count or does it actually copy the data?
[00:00:00 CET] --- Thu Jan  7 2016

More information about the Ffmpeg-devel-irc mailing list