[Ffmpeg-devel-irc] ffmpeg.log.20190909
burek
burek at teamnet.rs
Tue Sep 10 03:05:02 EEST 2019
[00:13:10 CEST] <hotbobby> looking for advice to get better quality out of nvenc. example command: ffmpeg -hwaccel nvdec -y -i input.mpeg2 -async 1 -f flv -c:v h264_nvenc -o rtmp://website.com:1935/stream
[00:13:38 CEST] <hotbobby> it will sit around 2000 kbit/s and looks pretty bad, but i also heard if you use a profile you can no longer use nvdec
[00:13:45 CEST] <hotbobby> just looking for any kind of advice, thanks
[00:13:58 CEST] <DHE> well you haven't specified any other options. maybe -b:v 6M or so?
[00:14:08 CEST] <DHE> how much bandwidth you got to spare?
[00:15:03 CEST] <DHE> also: ffmpeg -h encoder=h264_nvenc
[00:15:44 CEST] <hotbobby> i figured without setting a -crf it would use as much as it wanted
[00:15:54 CEST] <hotbobby> didnt know you could use help that way either, thanks
[00:16:04 CEST] <DHE> nope. otherwise "as much as it wanted" would be equivalent to -crf 0
[00:18:45 CEST] <hotbobby> looks like there are presets specific to h264_nvenc i can try too, thanks for all the help
[00:33:57 CEST] <hotbobby> looks like nvenc just looks bad at low bitrates :\
[00:34:20 CEST] <hotbobby> -preset slow -crf 0 will max out at like 3kbit/s from a 14kbit/s stream
[00:34:41 CEST] <hotbobby> input and output are both 720p60fps
[00:38:10 CEST] <another> 14kbit/s for 720p60 ?
[00:40:35 CEST] <hotbobby> 3000kbit/s and 14000kbit/s sorry
[00:43:13 CEST] <hotbobby> im transcoding video on a vm and want to preserve cpu usage. i figured nvdec/nvenc would be a good way to do so but there is no comparison to quality of x264
[00:53:45 CEST] <another> depends a bit on the hardware
[00:54:29 CEST] <another> recent nvidia cards are apparently quite good
[00:55:20 CEST] <another> but afaik they don't reach x264 in terms of compression efficiency
[01:08:00 CEST] <hotbobby> interesting, these are gtx970's ill have to try again with a current gen card
[01:23:32 CEST] <classsic> Hi, I have this commands https://gist.github.com/Classsic/49b00d4e68dc54a342f54a6cffb0823f and this inputs rtmp://207.246.92.193/copy/1 rtmp://207.246.92.193/copy/2 rtmp://207.246.92.193/copy/3 rtmp://207.246.92.193/copy/4rtmp://207.246.92.193/copy/1 rtmp://207.246.92.193/copy/2 rtmp://207.246.92.193/copy/3 rtmp://207.246.92.193/copy/4
[01:23:44 CEST] <classsic> sorry only 4 inputs rtmp://207.246.92.193/copy/1 rtmp://207.246.92.193/copy/2 rtmp://207.246.92.193/copy/3 rtmp://207.246.92.193/copy/4
[01:24:24 CEST] <classsic> the goal is get perfect sync, if somebody can helpme, or do it, I can pay some beers. :-)
[01:44:08 CEST] <markfilipak> ffprobe's calling params are documented, but how about the output? Specifically "23.98 fps, 23.98 tbr, 90k tbn, 47.95 tbc" ...frames per second, but what about the rest?
[01:49:12 CEST] <furq> markfilipak: container fps, best-guess codec fps, container timebase, codec timebase
[02:02:40 CEST] <markfilipak> furq: Thanks. Is that stuff documented anywhere?
[02:06:55 CEST] <markfilipak> Re, ffprobe: Is there any way to learn whether a stream is using constant frame rate or variable frame rate?
[02:07:04 CEST] <gtech> I have a digitized film that I didn't align correctly so the bottom of the next(previous?) frame appears at the top of the current frame. Is there a way to fix with ffmpeg?
[02:21:05 CEST] <gtech> https://imgur.com/a/b4SgqNe
[02:21:14 CEST] <gtech> There's an example of what I'm talking about.
[02:29:32 CEST] <nicolas17> gtech: if you copied that top strip into the bottom, would you get the full frame, or is there an extra missing gap?
[02:29:48 CEST] <nicolas17> guess I could try it
[02:30:05 CEST] <gtech> I mean it'd be a little better
[02:30:21 CEST] <gtech> You could do one more edit with removing the black strip
[02:30:53 CEST] <gtech> but yeah the idea is just to cut paste the top strip of video to the bottom.
[02:31:48 CEST] <nicolas17> I just tried and it doesn't fit at all, unless there's a lot of movement between those two frames?
[02:33:25 CEST] <gtech> Oh it must be overly zoomed too.
[02:33:30 CEST] <gtech> Derp
[02:33:39 CEST] <gtech> Then it's fine w/e.
[02:34:08 CEST] <gtech> Thanks for doing that :)
[04:28:27 CEST] <Diag> hey im likely about to ask a retarded question
[04:28:41 CEST] <Diag> but is there anything obvious that would prevent me from building ffmpeg on a ppc machine
[04:34:12 CEST] <furq> no
[04:36:40 CEST] <Diag> so i can just grab it and build like normal on linux?
[04:37:28 CEST] <furq> sure
[04:37:35 CEST] <Diag> hmm. ok
[04:37:47 CEST] <furq> the inline asm stuff is all optional and much of it probably exists on ppc anyway
[04:37:55 CEST] <Diag> oh nice
[04:38:00 CEST] <furq> maybe some external libs won't work, idk
[04:38:04 CEST] <furq> all the good ones should though
[04:38:07 CEST] <Diag> well, we'll see
[04:38:19 CEST] <Diag> I dont expect everything to just go. im mostly interested in just h264
[04:39:11 CEST] <furq> anyone competent will write the whole thing in c and then add asm fastpaths on top
[04:39:35 CEST] <furq> anyone competent who cares about other people using their library anyway
[04:40:04 CEST] <Diag> see thats the thin yaknow, ya'd hope that because its c, but...
[04:40:59 CEST] <furq> "anyone competent" being the operative term
[04:41:04 CEST] <furq> always important when you're talking about programmers
[04:41:37 CEST] <Diag> heh
[04:41:47 CEST] <Diag> random question by the way. has anyone built or ran ffmpeg on a ps3
[04:41:54 CEST] <Diag> totally uh, not related
[04:42:34 CEST] <furq> there's some specific asm in ffmpeg for cell iirc
[04:42:43 CEST] <furq> for optimisation, not to make it work
[04:42:48 CEST] <Diag> i just choked on my salad
[04:42:51 CEST] <Diag> theres WAHT
[04:43:04 CEST] <furq> idk if it made it in, i just remember seeing it on the ml years ago
[04:43:22 CEST] <Diag> *visibly excited* huh well uh, ok
[04:44:58 CEST] <furq> can't find it now but there's this
[04:45:00 CEST] <furq> https://github.com/FFmpeg/FFmpeg/blob/43b3412631893ca432857304e91d85b953bd5feb/configure#L5072
[04:45:13 CEST] <furq> so i guess it works just fine
[04:45:57 CEST] <Diag> i pooped a little
[04:46:42 CEST] <Diag> as soon as i finish dumping all the savedata off my ps3 ill drop linux on it and give er a whirl
[07:10:40 CEST] <Diag> havent done it yet
[07:10:42 CEST] <Diag> soon (tm)
[10:20:39 CEST] <YellowOnion> how do I burn a subtitle track in to my video?
[10:21:42 CEST] <YellowOnion> all the guides talk about using external files like a .srt file, but I don't have one.
[10:24:29 CEST] <furq> if the subs are in the file then use subtitles=input.mkv
[10:24:49 CEST] <furq> or if you have image subtitles then use scale2ref and overlay
[10:25:02 CEST] <YellowOnion> why can't I just tell it to grab the track directly?
[10:25:24 CEST] <JEEB> for image based subs you can, for text based nobody implemented that in ffmpeg.c yet
[10:25:47 CEST] <JEEB> that's why you pass the file (like, literally a file containing the subtitle stream) separately to the video filter
[10:26:55 CEST] <JEEB> video filters cannot take in AVSubtitles being one of the reasons why you can't just feed those subtitles to a video filter - unfortunately
[10:27:09 CEST] <JEEB> video and audio use AVFrames
[10:27:10 CEST] <YellowOnion> this is such a pain in the ass...
[10:27:41 CEST] <JEEB> do note that as I said, you can use the same video file that contains the suvtitle stream for the subtitle filter
[10:27:49 CEST] <JEEB> or you can attempt at making an extension to sub2video
[10:27:52 CEST] <JEEB> in ffmpeg.c
[10:28:14 CEST] <JEEB> that converts text based subtitles to images for the sub2video part to kick in
[10:28:57 CEST] <YellowOnion> it's basically impossible to script this...
[10:29:09 CEST] <JEEB> why?
[10:29:47 CEST] <YellowOnion> because \ is path seperator on windows.
[10:29:49 CEST] <JEEB> you can even specify in the subtitle filter which stream from the input file it is :P
[10:30:34 CEST] <JEEB> note: you can use / in Windows as well as far as I know, if that really really is a problem.
[10:31:20 CEST] <JEEB> it just means that in filter parameters you have to escape the backslash, just like with other things irrespective of the OS
[10:31:44 CEST] <JEEB> (I don't remember how much escaping you would have to do to be honest)
[10:31:54 CEST] <YellowOnion> either you add 3 backslashes or you replace \ with / still a pain ;)
[10:32:23 CEST] <JEEB> sure, but not especially "impossible to script"
[10:32:35 CEST] <JEEB> even with batch, although I feel sorry for you if that's what you've picked
[10:32:36 CEST] <JEEB> :)
[10:35:58 CEST] <YellowOnion> apparently this doesn't work because ffmpeg doesn't think it's a subtitle stream...
[10:43:22 CEST] <YellowOnion> so I guess the next question is how do I convert a mp4 "data" stream to a "sub" stream...
[10:44:25 CEST] <JEEB> by implementing the format whatever it may be?
[10:44:37 CEST] <JEEB> otherwise it's rather useless unless it can be interpreted
[10:44:53 CEST] <YellowOnion> Well it plays fine in VLC
[10:45:40 CEST] <JEEB> that still doesn't mean that it isn't just not implemented in VLC? VLC uses its own MP4 parser and it has support for various subtitle formats which people are keeping away from in FFmpeg due to XML etc
[10:46:03 CEST] <JEEB> ffprobe -v verbose -i INPUT , if you could post the output of that onto a pastebin or so and link here?
[10:47:05 CEST] <YellowOnion> Stream #0:4(und): Data: bin_data (text / 0x74786574)
[10:47:06 CEST] <YellowOnion> Metadata:
[10:47:06 CEST] <YellowOnion> creation_time : 2013-06-01T16:10:15.000000Z
[10:48:25 CEST] <JEEB> http://mp4ra.org/#/codecs
[10:48:30 CEST] <JEEB> doesn't seem to know "text"
[10:48:47 CEST] <JEEB> if VLC supports it, then its commit history might have hints
[10:50:24 CEST] <YellowOnion> I just wanted to bake these subs lol
[10:50:57 CEST] <JEEB> welcome to the real world where not everyone has poked your specific thingamajig
[10:51:29 CEST] <JEEB> #define ATOM_text VLC_FOURCC( 't', 'e', 'x', 't' )
[10:51:51 CEST] <JEEB> http://git.videolan.org/?p=vlc.git;a=blob;f=modules/demux/mp4/libmp4.c;h=a0beef3b9147ce5c406321f2e4394c2df1875990;hb=HEAD#l4707
[10:51:55 CEST] <JEEB> right
[10:52:08 CEST] <JEEB> it seems to just take it as an alias for the subt/tx3g/sbtl ones
[10:52:48 CEST] <YellowOnion> it's a "tx3g" sub format according to vlc.
[10:52:51 CEST] <JEEB> although in another nice list it's commented out? (´4@)
[10:53:00 CEST] <poutine> JEEB: Why is it 0x74786574 when that's literally "t","x","e","t"?
[10:53:22 CEST] <ritsuka> "text" is a mov fourcc
[10:53:32 CEST] <JEEB> poutine: they're written into the mov/mp4 container that way
[10:54:07 CEST] <JEEB> also ATOM_text being commented out here sounds like fun? http://git.videolan.org/?p=vlc.git;a=blob;f=modules/demux/mp4/libmp4.c;h=a0beef3b9147ce5c406321f2e4394c2df1875990;hb=HEAD#l4894
[10:54:38 CEST] <JEEB> but yes, if it's just another identifier for 3gpp/mov/mp4 timed text
[10:54:48 CEST] <JEEB> then just adding it into the mov list should work
[10:55:01 CEST] <JEEB> since we have support for that in FFmpeg as a subtitle format :P
[10:55:20 CEST] <YellowOnion> can I trick it in to using that without patching the code.
[10:57:46 CEST] <JEEB> by making a back-up of the file and binary patching with a hex editor? :P
[10:58:09 CEST] <YellowOnion> -tag:d? lol
[10:58:22 CEST] <JEEB> works for output I think :P
[10:58:28 CEST] <JEEB> if it works for inputs, sure :P
[10:58:34 CEST] <JEEB> but think that gets parsed from the container
[10:59:31 CEST] <JEEB> also if you mean remuxing I'm not sure if ffmpeg.c would keep the subtitle-related structures
[10:59:55 CEST] <JEEB> after all, it doesn't parse the media type or codec type correctly :P
[11:00:21 CEST] <JEEB> you could see with something like L-SMASH's boxdumper the byte offsets of the structures and see where the fourcc is located
[11:00:31 CEST] <JEEB> if you really want to rewrite the value
[11:00:52 CEST] <JEEB> `boxdumper --box FILE > FILE.box.log`
[11:09:11 CEST] <YellowOnion> according to mkvtoolnix it's a "chapters" stream and it no longer has subs if I remux it.
[11:10:17 CEST] <JEEB> because one of the ways chapters are passed on are as "subtitles" in mp4 :P
[11:10:19 CEST] <JEEB> or mov
[11:12:55 CEST] <YellowOnion> it would seem more sane to just extract the bitstream and the import it back with the correct fourcc?
[11:13:42 CEST] <JEEB> at this point if I knew it was timed text I'd probably just have added "text" into that one list and rebuilt FFmpeg :P
[11:14:02 CEST] <JEEB> if not possible, to make sure the file structure was not altered I would have used boxdumper + hex editor to modify a copy of the file
[11:14:21 CEST] <YellowOnion> on windows? lol
[11:14:39 CEST] <JEEB> yes?
[11:14:48 CEST] <JEEB> HxD is my go-to hex editor on windows after all
[11:15:10 CEST] <YellowOnion> I mean rebuilding ffmpeg on windows is not quick and painless.
[11:15:27 CEST] <JEEB> I dunno, native is slower than in a lunix vm, but nowadays you have WSL etc
[11:15:56 CEST] <YellowOnion> WSL? lol
[11:16:02 CEST] <YellowOnion> I think you're making my point for me.
[11:16:05 CEST] <JEEB> Windows Subsystem for Linux
[11:16:50 CEST] <JEEB> YellowOnion: I dunno what your definition of painless is. you can do it natively with a mingw-w64 compiler or MSVC, but then you gain the slowness of shell tools and I/O on windows (because of the antivirus and other reasons)
[11:17:12 CEST] <JEEB> and then if you just want to follow some ubuntu or whatever mingw-w64 toolchain tutorial, I mean WSL is there
[11:17:24 CEST] <JEEB> the first time setup is always the one part that bites you the most :P
[11:17:28 CEST] <JEEB> and anyways
[11:17:37 CEST] <JEEB> I did mention the hex editing if compilation is not possible
[11:18:08 CEST] <YellowOnion> yeah first time setup is what I mean...
[11:20:09 CEST] <YellowOnion> I tried it like a year ago to get opencl working
[11:20:44 CEST] <YellowOnion> apparently it was like one extra flag and some header files...but apparenly it's not the default still.
[11:22:16 CEST] <YellowOnion> I can probably just fine an srt file online too lol
[11:25:00 CEST] <JEEB> would have mostly been interesting to figure out if just interpreting the stream as such would make it work or not. but if you're not interesting in either patching & building OR hex editing a copy of the file (don't touch the original)
[12:39:10 CEST] <hero_biz> hi
[12:41:54 CEST] <hero_biz> guys, I'm trying to encode a video with x264 codec and resulting file gets too big. But I cannot see any place in video with high complexity with eye at least. I wonder if there is any way to profile complexity of video only? So I can find complex parts of my source which I cannot detect with my eye?
[12:53:08 CEST] <BtbN> Use the segment muxer and look at the biggest segments
[13:20:53 CEST] <Radiator> Hey all, does libav* have any functions to do any gui programming like opening a window and so on ?
[13:21:25 CEST] <ritsuka> no
[13:22:03 CEST] <Radiator> Ok, what do you use for the GUI ? I've used opencv2 but it isn't as intuitive as I expected.
[13:25:06 CEST] <DHE> normally GUI functions are part of the OS itself. ffmpeg provides video support
[13:26:06 CEST] <Radiator> That's what I thought
[13:34:26 CEST] <JEEB> Radiator: generally it works the other way. you have a GUI application and then you utilize the FFmpeg libraries to come up with something to render onto that view
[14:52:56 CEST] <^Neo> Good morning! Has anyone used a Sencore SDI2X to convert SDI to SMPTE 2110-20/2022-6?
[15:12:38 CEST] <Radiator> Is it possible to use CUDA or VAAPI to decode ?
[15:45:55 CEST] <^Neo> Radiator - https://trac.ffmpeg.org/wiki/HWAccelIntro
[15:49:28 CEST] <BtbN> nvidia does not support nvdec. It's either vdpau or nvdec there.
[15:49:43 CEST] <BtbN> You cannot use CUDA itself for decoding, but it's supporting API, nvdec.
[15:49:46 CEST] <Radiator> ^Neo: Thanks :)
[15:49:59 CEST] <BtbN> *nvidia does not support _vaapi_
[15:50:12 CEST] <Radiator> BtbN: Yeah sorry that's what I meant by saying CUDA
[15:50:41 CEST] <Radiator> It's either Intel or Nvidia anyway
[15:51:31 CEST] <Radiator> vaapi supporting Intel and nvdec nvidia, right ?
[15:51:57 CEST] <BtbN> For just decoding and playing there is no difference between vdpau and nvdec, and vdpau has better player support.
[15:52:11 CEST] <BtbN> If this is an Optimus-Laptop, keep in mind that the nvidia GPU most likely has no video de/encoding capabilities.
[15:52:26 CEST] <Radiator> BtbN: What do you mean by playing ?
[15:52:57 CEST] <BtbN> Well, playing the video?
[15:54:05 CEST] <Radiator> Beside decoding a stream to retreive the frame i don't see want to can do more to play the video
[15:54:11 CEST] <BtbN> ?
[15:54:20 CEST] <BtbN> You play it. To watch it. What else do you want to do with it?
[15:54:24 CEST] <Radiator> *I don't see what you can do
[15:55:07 CEST] <Radiator> Oooh sorry my mind is stuck on the library not on the ffmpeg or avconv binaries which display the video
[15:55:11 CEST] <Radiator> My bad
[15:55:20 CEST] <BtbN> They don't oO
[15:55:29 CEST] <BtbN> They're cli tools
[15:55:40 CEST] <BtbN> Any video PLAYER, that PLAYS the video, makes use of those APIs
[15:56:13 CEST] <BtbN> vdpau and vaapi come with a full presentation chain. nvdec does not, unless you do some trickers with mapping things to GL/Vulkan
[15:56:45 CEST] <Radiator> I didn't know !
[15:57:12 CEST] <Radiator> I still have a lot to learn about those APIs
[17:53:32 CEST] <thnee> I am trying to play an HLS stream with ffplay. It says "Failed to open segment", even though the file works fine? https://dpaste.de/byYJ
[17:53:46 CEST] <thnee> I have also verified downloading it with curl and playing it that way, which works just fine.
[17:54:20 CEST] <thnee> Even with -logleve trace, ffplay does not provide any information about *why* it thinks it has failed, please help?
[17:59:56 CEST] <poutine> thnee: says 0B
[18:00:40 CEST] <poutine> [http @ 0x7fa4b8012700] header='HTTP/1.1 200 OK'B sq= 0B f=0/0
[18:39:24 CEST] <thnee> Umm ok..
[18:40:09 CEST] <thnee> poutine: Thanks, but what the heck does that mean?
[18:49:24 CEST] <poutine> thnee: sorry I checked the source code and that's actually a reference to something relating to subtitles, ignore my comment
[20:17:19 CEST] <leif> Hey, this is probably not the right place to ask, but I don't really know of anywhere better (nor do I really know the right way to ask it):
[20:17:52 CEST] <leif> Is there any cheap device that does nothing but forwards an hdmi signal, and doesn't support HDCP.
[20:18:13 CEST] <leif> (As in, if a laptop asks if it supports hdcp, it says no.)
[20:18:52 CEST] <leif> I ask because apparently mac's like to send an hdcp signal if the target device can support it.
[20:19:12 CEST] <leif> Which makes are splitter (and capture ffmpeg script) not work.
[20:19:26 CEST] <leif> (For reference, we're just trying to capture presentor slides)
[20:19:53 CEST] <leif> Does anyone have any thoughts, or even know of the right place to ask that question?
[21:19:43 CEST] <Retal> Hello guys
[21:19:43 CEST] <Retal> I am trying get info about UDP stream bitrate info. I see frame size, fps.....
[21:19:43 CEST] <Retal> but no bitrate: bitrate: N/A
[21:21:15 CEST] <Retal> how to find bitrate info for UDP streams ?
[21:47:41 CEST] <SimAV> Hi! I would like ffmpeg to draw a *precise* waveform, i.e. exactly for every video frame the audio that should be played during the presentation of that frame
[21:48:33 CEST] <SimAV> I tried with https://ffmpeg.org/ffmpeg-filters.html#showwaves, however I did not get it output one drawn frame per video frame...
[21:50:41 CEST] <SimAV> I tried e.g.: ffmpeg -i INPUT.mkv -filter_complex '[0:a] showwaves=r=30 [A]; [0:v] [A] overlay [V] ' -map '0:a' -map '[V]' -b:v 28M -c:v h264_nvenc -f mp4 OUTPUT.mp4
[21:51:17 CEST] <SimAV> as ffprobe INPUT.mkv outputs
[21:51:18 CEST] <SimAV> Stream #0:0(eng): Video: vp9 (Profile 0), yuv420p(tv, bt709), 1920x1080, SAR 1:1 DAR 16:9, 30 fps, 30 tbr, 1k tbn, 1k tbc (default)
[21:53:38 CEST] <SimAV> but this doesn't deliver even just plausible results... (e.g., some waveform-frames seem to be just a copy of the preceding waveform-frame)
[21:54:55 CEST] <SimAV> Does someone in here have an idea how to achieve this? (not necessarily, but preferably using ffmpeg)
[21:55:29 CEST] <Atlenohen> Hello
[21:55:55 CEST] <Atlenohen> is this still needed for HLS downloads "-bsf:a aac_adtstoasc"
[21:56:05 CEST] <Atlenohen> I opened in MVP and it was able to play audio
[21:56:16 CEST] <Atlenohen> I guess it's just some compatability with something else, but works on desktop
[21:56:49 CEST] <Atlenohen> Oh, I downloaded raw without conversion, it did play raw .TS AAC ADTS normally
[21:58:40 CEST] <durandal_1707> SimAV: you know that you would get dupe waveform anyway as that is result of big resolution and frame rate
[21:59:48 CEST] <durandal_1707> reduce width of waveform video frame to get exactly unique waveforms each frame
[21:59:52 CEST] <SimAV> durandal_1707, what do you mean by dupe waveform? some part of the 'audio frame' being plottet twice?
[21:59:59 CEST] <durandal_1707> yes
[22:00:27 CEST] <SimAV> Stream #0:1(eng): Audio: aac (LC), 44100 Hz, stereo, fltp (default)
[22:00:27 CEST] <durandal_1707> or resample audio to higher sample rate
[22:00:38 CEST] <SimAV> so I tried with
[22:00:40 CEST] <SimAV> -filter_complex '[0:a] showwaves=mode=p2p:scale=lin:split_channels=0:s=1470x1080:r=30 [A]; [0:v] [A] overlay [V] '
[22:00:55 CEST] <SimAV> as 44100/30 = 1470
[22:01:12 CEST] <SimAV> durandal_1707, but that didn't solve the problem either...
[22:02:33 CEST] <SimAV> again, some frames show exactly the same waveform
[22:03:43 CEST] <durandal_1707> is overlay video 30 fps?
[22:04:26 CEST] <SimAV> as posted above, Stream #0:0 (this is the video) has 30 fps
[22:04:59 CEST] <SimAV> and Stream #0:1 (this is the audio) has 44100Hz as samplerate
[22:05:37 CEST] <durandal_1707> right. let me try something
[22:11:32 CEST] <SimAV> (just in case this should be relevant: I'm using ffmpeg version 4.1.4 on debian buster)
[22:13:07 CEST] <durandal_1707> can not reproduce it
[22:13:49 CEST] <durandal_1707> with 5.1 48000 audio and higher resolution
[22:13:51 CEST] <SimAV> durandal_1707, what command did you use?
[22:14:20 CEST] <durandal_1707> my video is not 30 but 24
[22:15:00 CEST] <durandal_1707> i just added fps=30 for video input
[22:16:25 CEST] <durandal_1707> im using git version of ffmpeg
[22:19:39 CEST] <durandal_1707> the reason you get same frame when overlaying is probably because of overlay filter
[22:56:41 CEST] <SimAV> durandal_1707, OK, If i omit the overlay and just render the waveform video, that result seems more reasonable
[22:57:55 CEST] <SimAV> if I colorkey the waveform in a second ffmpeg invocation and overlay it over the original video there, it seems as if the wave-form is one frame late
[22:59:45 CEST] <durandal_1707> the output of waveform filter always start at 0
[23:00:24 CEST] <durandal_1707> the audio encoder you use may add additional delay
[23:01:07 CEST] <durandal_1707> so for intermediate files use lossless audio and video codecs
[23:01:18 CEST] <SimAV> durandal_1707, no, the audio tracks are exactly equivalent
[23:01:45 CEST] <SimAV> the overlayed waveform ends up one frame later than it should
[23:02:31 CEST] <durandal_1707> open bug report with exact steps to reproduce issue
[23:03:35 CEST] <SimAV> so in the pure waveform video frame 70 (which shows the audio during original video frame 70) is overlayed over original video frame 71
[23:04:10 CEST] <SimAV> (when using two ffmpeg invocations, first creating the pure waveform video, second colorkeying and overlaying it)
[23:04:33 CEST] <SimAV> durandal_1707, Ok, I have to get an example video first, that I can publish...
[23:05:15 CEST] <durandal_1707> anyway try latest ffmpeg version
[23:09:23 CEST] <SimAV> durandal_1707, ok. I'll do that and be back tomorrow... Thanks!
[23:36:28 CEST] <SimAV> durandal_1707, Ok, I just reproduced the problem with current git (558265d4bc)
[23:36:51 CEST] <SimAV> anyways, I'll be back tomorrow...
[00:00:00 CEST] --- Tue Sep 10 2019
More information about the Ffmpeg-devel-irc
mailing list