[Ffmpeg-devel-irc] ffmpeg.log.20150919

burek burek021 at gmail.com
Sun Sep 20 02:05:01 CEST 2015


[00:21:47 CEST] <ewew> hello
[00:21:47 CEST] <ewew> can i ask about avconf ?
[00:22:37 CEST] <ewew> ups avconv
[00:22:56 CEST] <c_14> No, either use ffmpeg (from FFmpeg) or ask in #libav
[00:22:58 CEST] <durandal_1707> no, this is about ffmpeg
[00:29:40 CEST] <kapusta> hi there :)
[00:30:18 CEST] <kapusta> anyone have an idea how to encode an mpeg1video using two seperate key and fill targa sequences?
[00:31:21 CEST] <kapusta> got RGB as one sequence and Alpha as separate sequence but when I encode using two inputs and set complex filter to overlay the alpha channel ain't transparent
[00:33:12 CEST] <durandal_1707> perhaps you need alphamerge
[00:33:44 CEST] <kapusta> i'm listening ;)
[00:34:33 CEST] <durandal_1707> There are docs
[00:35:01 CEST] <kapusta> ah ok, i'll try and google this up
[01:23:15 CEST] <waressearcher2> this is weird
[01:24:59 CEST] <waressearcher2> believe it or not but these two commands make two different videos: "ffmpeg -i video.avi -i audio.ac3 -c:v copy -map 0:v:0 -map 1:a:0 -shortest -y -f avi out.avi",   "ffmpeg -i audio.ac3 -i video.avi -c:v copy -map 0:a:0 -map 1:v:0 -shortest -y -f avi out.avi" ?
[01:25:03 CEST] <waressearcher2> why ?
[01:25:57 CEST] <waressearcher2> for a better comparison
[01:26:11 CEST] <waressearcher2> ffmpeg -i video.avi -i audio.ac3 -c:v copy -map 0:v:0 -map 1:a:0 -shortest -y -f avi out.avi
[01:26:14 CEST] <waressearcher2> ffmpeg -i audio.ac3 -i video.avi -c:v copy -map 0:a:0 -map 1:v:0 -shortest -y -f avi out.avi
[01:27:36 CEST] <waressearcher2> few hours ago or even day ago I was asking few times about my problem with adding silent audio and all I had to do just exchange parameters in command, after that I got no errors
[01:29:46 CEST] <waressearcher2> no wait
[01:31:27 CEST] <waressearcher2> to be exact, these two commands:
[01:32:27 CEST] <waressearcher2> ffmpeg -i video.avi -i audio.ac3 -c:v copy -map 0:v:0 -map 1:a:0 -shortest -y -f avi out.avi
[01:32:31 CEST] <waressearcher2> ffmpeg -i audio.ac3 video.avi -c:v copy -map 0:a:0 -map 1:v:0 -shortest -y -f avi out.avi
[01:32:40 CEST] <waressearcher2> in both cases I checked md5sum and those files are different
[01:33:14 CEST] <waressearcher2> isn't it weird ?
[01:33:27 CEST] <waressearcher2> I use ffmpeg-2.3
[01:33:49 CEST] <waressearcher2> I mean it can't be "failure" of ffmpeg, but what is the explanation ?
[02:22:33 CEST] <BullHorn> hello
[02:23:10 CEST] <BullHorn> i recorded a video into .flv format (streaming) and now im trying to change the container to .mp4 for video editing
[02:23:16 CEST] <BullHorn> i run
[02:23:16 CEST] <BullHorn> ffmpeg -i D:\Streaming\nvenc.flv -c copy -copyts F:\testHUGE.mp4
[02:23:22 CEST] <BullHorn> and i get
[02:23:23 CEST] <BullHorn> [flv @ 0000000004e78280] Packet mismatch 347 344
[02:23:33 CEST] <BullHorn> any ideas? it used to work in the past, im not sure whats gone crazy
[02:30:49 CEST] <cbsrobot_> waressearcher2: well in the first case you have: stream 0: video, stream 1: audio
[02:31:05 CEST] <cbsrobot_> in the second: stream 0: audio, stream 1; video
[04:05:44 CEST] <DHE> I have a video that's a mixture of regular interlacing and telecine'd film at 1080i 30fps source. is there a good way to deal with this and produce idealized progressive video? ffmpeg-filters suggests not really. ideally I'd want the delay between frames is constant (up to switching between modes) and can tolerate an FPS change...
[04:05:50 CEST] <DHE> but I suspect I'm asking too much
[04:44:45 CEST] <durandal_1707> DHE: tried pullup?
[07:44:34 CEST] <zylthinking> avctx->extradata have 5 bytes for aac,  0x15, 0x08, 0x56, 0xe5, 0xb8 what these are? seems the first 2 bytes are audio specific config, but the latter 3 bytes?
[10:44:02 CEST] <dv_> I am writing a library for accessing a hardware video encoder which can encode h264, motion JPEG, mpeg4part2, h263. with mpeg4/h264/mjpeg, I have to manually tell the encoder to add headers (VOS/VIS/VOL headers for mpeg4, SPS/PPS for h264, JPEG headers for MJPEG). now I wonder what the recommendation is, when to add them.
[10:44:17 CEST] <dv_> at each I/IDR frame? in certain intervals? on every frame?
[10:45:07 CEST] <dv_> I can make this configurable of course, but even then I need some defaults. I mention this here both because of the know-how you guys have and because this library might be used in ffmpeg some day.
[11:15:24 CEST] <Mavrik> dv_, the default is usually to attach them before each I/IDR frame
[11:15:31 CEST] <Mavrik> that's what x264 does
[11:16:22 CEST] <Mavrik> Let's you start playing video immediately on a keyframe and doesn't make sense anywhere else since you can't start from any other point anyway.
[11:17:26 CEST] <dv_> yeah, thats what I thought. same goes for mpeg4p2?
[11:17:50 CEST] <dv_> but for mjpeg, I think I should add headers to every frame, since mjpeg is effectively just a bunch of concatenated jpegs, no?
[11:18:21 CEST] <Mavrik> Does mjpeg even have headers?
[11:18:30 CEST] <dv_> well, the regular jpeg headers
[11:19:13 CEST] <dv_> oh, but what about intra refresh mode in h264? you dont have idr/i with this, do you
[11:20:27 CEST] <Mavrik> You still have an interval on which you get a full frame.
[11:22:21 CEST] <dv_> hm, I dont get it. isnt intra refresh supposed to be an alternative to the I/IDR mechanism? that you are guaranteed to get a reconstructed frame after N slices?
[11:22:44 CEST] <dv_> how then can you have intervals with "full" frames?
[11:22:59 CEST] <dv_> or do you refer to these N slices?
[11:23:48 CEST] <Mavrik> Yes
[11:24:09 CEST] <Mavrik> You do need to assemble a full frame eventually - that being a GOP slice :)
[11:24:45 CEST] <dv_> yeah, but these intervals arent placed in a fixed position. if you start at position P - *any* position P - you get a reconstructed frame after P+N slices.
[11:25:00 CEST] <dv_> so I wouldnt know where to add SPS/PPS headers.
[11:27:04 CEST] <Mavrik> You can put them anywhere.
[11:27:14 CEST] <Mavrik> Before the player can start it has to read N slices.
[11:27:28 CEST] <Mavrik> It's ideal that in that interval it also encounters SPS/PSS to start as soon as possible.
[11:29:28 CEST] <dv_> right. so event if they are in the middle of this N-slice-interval, the player still sees them
[11:29:57 CEST] <dv_> the interval just has to conform to the "GOP size", which is essentially the N value. I know that "GOP" doesnt really exist as a concept in h264 anymore.
[11:30:33 CEST] <Mavrik> Well, it doesn't have to conform really - they can be injected even on greater interval.
[11:30:52 CEST] <dv_> sure, but I want to make sure this stuff also works nicely for streams
[11:30:52 CEST] <Mavrik> But the player can't start playing until it encounters those parameters so if they're too rare, seeking is annoying.
[11:31:02 CEST] <dv_> and that too
[12:44:41 CEST] <waressearcher2> cbsrobot_: in the first case I have "-i video.avi -i audio.ac3" and I use "-map 0:v:0 -map 1:a:0" so it matches, in the second case I have "-i audio.ac3 -i video.avi" and I use "-map 0:a:0 -map 1:v:0" I reverse both, so it matches again, but still output files are differ
[12:54:39 CEST] <cbsrobot_> waressearcher2: the postion of the -map in your command line is important
[12:54:53 CEST] <cbsrobot_> first -map = stream 0
[12:54:56 CEST] <BullHorn> hello
[12:54:59 CEST] <BullHorn> i recorded a video into .flv format (streaming) and now im trying to change the container to .mp4 for video editing
[12:55:02 CEST] <BullHorn> i run
[12:55:02 CEST] <cbsrobot_> second -map = stream 1
[12:55:04 CEST] <BullHorn> ffmpeg -i D:\Streaming\nvenc.flv -c copy -copyts F:\testHUGE.mp4
[12:55:07 CEST] <BullHorn> and i get
[12:55:09 CEST] <BullHorn> [flv @ 0000000004e78280] Packet mismatch 347 344
[12:55:11 CEST] <BullHorn> any ideas? it used to work in the past, im not sure whats gone crazy
[12:56:50 CEST] <waressearcher2> cbsrobot_: in both cases "-map 0:v:0" points to "-i video.avi" and "-map 1:a:0" to "-i audio.ac3"
[12:58:44 CEST] <waressearcher2> cbsrobot_: but videos are allright if you play them standalone, the thing is if I concatenate video from first command with two other video and put it in between, there goes the difference, when mplayer starts to play that concatenated video in first case it is normal in second case it stops playing when it reaches that part in side the video,
[12:58:47 CEST] <cbsrobot_> yeah - try: first case: -map 0:v:0 -map 1:a:0
[12:58:49 CEST] <waressearcher2>  that all I can write as explanation now, because I spend 2 days here explaining that here no one seems to understand
[12:58:59 CEST] <cbsrobot_> second case: -map 1:v:0 -map 0:a:0
[12:59:14 CEST] <waressearcher2> cbsrobot_: videos are all right on their own, they compeletely allright the problem goes when I concatenate them
[12:59:32 CEST] <waressearcher2> thats is the weird thing, they are both compeletely allright
[13:00:09 CEST] <cbsrobot_> the order of the -map (video) and -map (audio) should be the same
[13:00:21 CEST] <cbsrobot_> so always put -map (select video stream) first
[13:00:34 CEST] <cbsrobot_> and then -map (select audio stream)
[13:01:25 CEST] <waressearcher2> cbsrobot_: so which of these two command do you think is correct ?
[13:01:31 CEST] <waressearcher2> ffmpeg -i video.avi -i audio.ac3 -c:v copy -map 0:v:0 -map 1:a:0 -shortest -y -f avi out.avi
[13:01:34 CEST] <waressearcher2> ffmpeg -i audio.ac3 video.avi -c:v copy -map 0:a:0 -map 1:v:0 -shortest -y -f avi out.avi
[13:02:00 CEST] <cbsrobot_> the first command will put stream0: video, stream1: audio
[13:02:09 CEST] <cbsrobot_> the first command will put stream0: audio, stream1: video
[13:02:13 CEST] <waressearcher2> and second
[13:02:14 CEST] <cbsrobot_> see the difference ?
[13:02:37 CEST] <cbsrobot_> try these:
[13:02:46 CEST] <cbsrobot_> A) ffmpeg -i video.avi -i audio.ac3 -c:v copy -map 0:v:0 -map 1:a:0 -shortest -y -f avi out.avi
[13:03:04 CEST] <cbsrobot_> B) ffmpeg -i audio.ac3 video.avi -c:v copy -map 1:v:0 -map 0:a:0 -shortest -y -f avi out.avi
[13:03:35 CEST] <waressearcher2> cbsrobot_: right when I do ffprobe on those videos for first video it gaves: "Stream #0:0: Audio" and for second "Stream #0:0: Video" but why is it matter in what order are streams inside the file ?
[13:03:48 CEST] <cbsrobot_> because
[13:04:15 CEST] <cbsrobot_> read the concatenate manual again
[13:04:23 CEST] <waressearcher2> because it will matter when concatenate them ?
[13:04:54 CEST] <cbsrobot_> https://ffmpeg.org/ffmpeg-formats.html#concat , 4th paragraph
[13:05:06 CEST] <cbsrobot_> All files must have the same streams (same codecs, same time base, etc.).
[13:05:21 CEST] <cbsrobot_> waressearcher2: yes
[13:11:58 CEST] <Mavrik> what's the current status of AAC vs. FDK_AAC?
[13:13:58 CEST] <waressearcher2> cbsrobot_: but still why can't ffmpeg just recognise the right order itself
[13:22:18 CEST] <waressearcher2> cbsrobot_: also why don't ffmpeg just place always video stream first and audio second, that way order always will be the same and there will be no problems, I mean why allow to change order, who needs to change it anyway
[13:36:40 CEST] <cbsrobot_> waressearcher2: every user has it's own needs - that is yours
[13:36:54 CEST] <cbsrobot_> but I guess there are a gazillion of other usescases
[13:37:22 CEST] <cbsrobot_> like one video and 5 different languages and 31 different subtitle streams
[13:38:01 CEST] <DHE> Mavrik: pretty sure the opinion is fdk is the best choice
[13:38:25 CEST] <cbsrobot_> good luck recognising the right order in this case
[13:39:13 CEST] <Mavrik> DHE, I'm aware of that, my question is more - is AAC now good enough for production use
[13:39:16 CEST] <cbsrobot_> Mavrik: the insternal aac encoder got a lot of imporvements and I guess the experimental flag was removed
[13:39:41 CEST] <Mavrik> ah.
[13:40:09 CEST] <cbsrobot_> see ticket#2686 for the detailed status
[13:44:42 CEST] <Mavrik> Mhm, am following it. Will test it on our own encodes, kinda hoping to get rid of that external dependency :)
[14:44:39 CEST] <waressearcher2> cbsrobot_: thanks
[14:51:07 CEST] <iive> DHE: i think your info might be slightly outdated :D
[14:52:45 CEST] <DHE> oh? stock AAC has improved that much?
[14:53:58 CEST] <waressearcher2> why do you care about aac so much, why not to use ogg ?
[14:54:12 CEST] <waressearcher2> also why is it such a problem to create good aac codec ?
[14:54:32 CEST] <Mavrik> ...
[14:56:12 CEST] <DHE> why choose a codec, some people are working with hardware or software with limitations
[14:56:33 CEST] <DHE> as for codec creation, even Vorbis has gone through a few fairly significant code changes over the years in the name of quality
[14:59:05 CEST] <Mavrik> I'm sure the DVB equipment will just love streams with ogg
[15:00:40 CEST] <iive> ogg is the file format, vorbis is the audio codec
[16:02:09 CEST] <waressearcher2> iive: and someone called you "ignorant"
[16:02:14 CEST] <waressearcher2> iive: day ago
[16:03:16 CEST] <iive> let's just assume they had a bad day :)
[16:27:09 CEST] <luc4> Hello! I read that if I update ffmpeg in my working app from 2.a to 2.b with b > a I should keep getting a working app. Is this correct?
[16:29:33 CEST] <waressearcher2> that called backward compatibility
[16:29:45 CEST] <waressearcher2> whatis your "working app" ?
[16:30:19 CEST] <waressearcher2> at least it "should" work
[16:30:33 CEST] <waressearcher2> can be some specifics related to what app exactly it is
[16:33:10 CEST] <waressearcher2> thats as fuzzy answer as one give from as fuzzy question as one get
[16:38:57 CEST] <luc4> waressearcher2: I wrote a complex library that uses ffmpeg to open containers and uses hardware acceleration to decode and render using opengl. The implementation is complex, but I just noticed that when switching from 2.6 (or probably 2.7.2) to 2.8, make it impossible to decode for some unknown reason. Im trying to determine what exactly is different, I just wanted a confirmation.
[16:39:28 CEST] <luc4> It seems to happen on a specific video file.
[16:39:46 CEST] <luc4> (or category of course, Im experiencing on one specific video file)
[16:43:39 CEST] <Mavrik> Uh, no, API isn't kept deliberately compatible between 2.x releases.
[16:43:43 CEST] <Mavrik> Neither is ABI.
[16:44:06 CEST] <luc4> Mavrik: I was referring to this: https://www.ffmpeg.org/doxygen/2.5/index.html
[16:44:31 CEST] <luc4> Mavrik: In other words, any correct program that works with a given FFmpeg snapshot should work just as well without any changes with any later snapshot with the same major versions.
[16:44:52 CEST] <BtbN> It's not talking about the ffmpeg version number.
[16:45:01 CEST] <BtbN> It's talking about the individual library version numbers.
[16:45:12 CEST] <BtbN> But there wasn't a major bump between 2.7 and 2.8
[16:46:26 CEST] <luc4> Ah I see& so there may be something that makes this happen&
[16:47:17 CEST] <BtbN> If you aren't running into any deprecation warnings, there should be nothing too major though.
[16:47:42 CEST] <luc4> Unfortunately I couldnt determine what was wrong with 2.8& everything seemed fine. Still seems to be working properly with 2.6 and not working with that file with 2.8.
[16:47:58 CEST] <luc4> Im rebuilding using 2.7.2 and Ill see what happens with that.
[16:49:04 CEST] <luc4> What seems particular of that file is that there is no audio and that there is a data stream.
[16:49:50 CEST] <luc4> However I see that 2.8 properly extracts video frames, provides those to me and I provide those to the decoder. The decoder is not sending me decoded frames.
[16:50:21 CEST] <BullHorn> BtbN :3
[16:51:11 CEST] <BullHorn> im still having issues but im not sure its caused by ffmpeg anymore
[16:51:21 CEST] <BullHorn> i record video in OBS into .flv container using NVENC
[16:51:31 CEST] <BullHorn> ffmpeg is unable to convert the container to .mp4 like it used to when i had win7
[16:51:42 CEST] <BullHorn> looking in MediaInfo looks like something is not done right, probably in OBS :(
[16:55:28 CEST] <luc4> BtbN: 2.7.2 works properly
[16:56:37 CEST] <luc4> BtbN: oh, there was something weird about 2.8 that I noticed& it was failing to build with my 2.7 configure line. It required h263 decoder.
[16:57:04 CEST] <BtbN> Then check https://github.com/FFmpeg/FFmpeg/blob/master/doc/APIchanges#L41 for a list of api changes in 2.8
[16:57:26 CEST] <BtbN> Assuming you are only using public api, that's all you have to look out for.
[16:58:22 CEST] <luc4> BtbN: ok, Ill check this. Thanks.
[16:59:43 CEST] <luc4> BtbN: I dont see this in there, but is now h263 mandatory?
[17:00:10 CEST] <BtbN> what?
[17:01:24 CEST] <waressearcher2> it can be disabled by --disable-h263 right ?
[17:01:36 CEST] <luc4> Yes, that is the question.
[17:01:44 CEST] <BtbN> What is the question?
[17:01:56 CEST] <luc4> Is h263 now mandatory in 2.8?
[17:02:04 CEST] <BtbN> I have no idea what you mean by that.
[17:03:19 CEST] <luc4> I have a script that builds ffmpeg for my arm platform. I keep updating the ffmpeg version leaving the rest unchanged. When I changed to 2.8 I got these: http://pastebin.com/uYkYiLb7.
[17:03:42 CEST] <luc4> In my script h263 is disabled.
[17:04:00 CEST] <BtbN> With what configure line?
[17:04:10 CEST] <luc4> I enabled it after reading the errors, and it builds properly now.
[17:04:23 CEST] <BullHorn> help http://i.imgur.com/zsXznZI.png
[17:04:26 CEST] <BtbN> Sounds like a bug to me, do you have a minimal configure line that reproduces it?
[17:04:28 CEST] <BullHorn> what is the meaning of this
[17:04:33 CEST] <luc4> BtbN: https://github.com/carlonluca/pi/blob/master/piomxtextures_tools/compile_ffmpeg.sh
[17:04:55 CEST] <BtbN> BullHorn, it means you should increase the analyzeduration.
[17:05:20 CEST] <luc4> BtbN: unfortunately it is for a arm device, not sure the same happens on desktop, probably...
[17:05:30 CEST] <luc4> BtbN: it should
[17:05:40 CEST] <BtbN> That's nearly 200 lines of configure. Something more minimal would be great.
[17:06:29 CEST] <BullHorn> BtbN what would i need to add here to raise the analyzeduration and probesize: ffmpeg -i F:\input.flv -c copy -copyts F:\output.mp4
[17:06:32 CEST] <BtbN> You're settings a huge bunch of options, it's quite possible that combination was never tested and triggers some obscure bug.
[17:06:44 CEST] <BtbN> Well, analyzeduration and probesize?
[17:06:59 CEST] <BullHorn> ye but how, i have no idea about the syntax
[17:07:11 CEST] <BtbN> -analyzeduration and -probesize
[17:07:30 CEST] <BtbN> + apropriate values
[17:07:37 CEST] <BullHorn> what are teh default
[17:07:44 CEST] <BtbN> No idea, check -help full
[17:08:57 CEST] <luc4> BtbN: yes, of course it is possible. Ill see if I can create something smaller.
[17:09:50 CEST] <BtbN> I'd definitely open a bug about that, even if you don't manage to find a more minimal configure line.
[17:09:50 CEST] <BullHorn> i raised both analyzeduration and probesize to 100000000000 and getting the same error
[17:09:52 CEST] <BullHorn> :(
[17:10:01 CEST] <BtbN> Can you upload the file somewhere?
[17:10:07 CEST] <BullHorn> sure
[17:10:08 CEST] <BullHorn> sec
[17:10:49 CEST] <BullHorn> https://mega.nz/#!2wMzEAwT!wNYysvN1Gqk62BH2CEOeI5KJ534CjekmHfeKm3iMQ5g
[17:11:36 CEST] <BtbN> That's a very small file, is that right?
[17:11:39 CEST] <BullHorn> yes
[17:11:41 CEST] <BullHorn> 5sec test
[17:11:49 CEST] <BullHorn> but its the same for any other .flv that i record in OBS
[17:11:57 CEST] <BtbN> https://bpaste.net/show/50c502477256
[17:12:27 CEST] <BullHorn> lines 20-21
[17:12:31 CEST] <BullHorn> my ffmpeg doesnt recognize this
[17:12:33 CEST] <BullHorn> why?
[17:12:40 CEST] <BullHorn> im using latest x64 windows ffmpeg build
[17:12:58 CEST] <BtbN> Remuxing to mp4 also works fine for me.
[17:13:13 CEST] <BtbN> No idea, i'll try again with current master.
[17:13:25 CEST] <luc4> BtbN: thanks
[17:19:54 CEST] <BullHorn> i deleted the old build i had before
[17:21:43 CEST] <BullHorn> or maybe im just missing codecs?
[17:21:52 CEST] <BullHorn> or does ffmpeg come with all thats necessary for that file format
[17:23:45 CEST] <BullHorn> found anything interesting BtbN?
[17:27:00 CEST] <BtbN> Just done compiling
[17:27:12 CEST] <BullHorn> oh wait, i need to complile it?
[17:27:21 CEST] <BullHorn> i just took the latest static .zip, unpacked and.. yeah
[17:27:21 CEST] <BullHorn> ;x
[17:29:14 CEST] <BtbN> BullHorn, yes, i can confirm, latest master has issues with that file.
[17:29:20 CEST] <BtbN> 2.8 works fine
[17:29:28 CEST] <BullHorn> cool, where do they keep older builds
[17:29:35 CEST] <BtbN> they?
[17:29:40 CEST] <BullHorn> ffmpeg guys
[17:29:40 CEST] <BullHorn> ;x
[17:29:43 CEST] <BtbN> ffmpeg does not keep builds anywhere, it just provides the sources.
[17:41:34 CEST] <BullHorn> the build from september 14th works
[17:42:02 CEST] <BullHorn> BtbN are you registered on ffmpeg website? can you report the bug?
[17:42:33 CEST] <BtbN> do you have the git hash of that build?
[17:42:47 CEST] <BullHorn> https://ffmpeg.zeranoe.com/builds/win64/static/
[17:42:47 CEST] <BtbN> It's in the very first line it outputs
[17:42:49 CEST] <BullHorn> this one works: ffmpeg-20150915-git-8e92462-win64-static.7z
[17:42:50 CEST] <gromit> Anybody knows my testing repository, made with ffmpeg: http://av-standard.irt.de/wiki/index.php/Referenzclips
[17:42:51 CEST] <BullHorn> anything after it doesnt
[17:43:06 CEST] <BullHorn> ffmpeg version N-75287-g8e92462 Copyright (c) 2000-2015 the FFmpeg developers
[17:44:18 CEST] <BullHorn> hmmm its converting so slowly wtf
[17:44:27 CEST] <BtbN> You did add -c copy?
[17:44:29 CEST] <BullHorn> only slightly faster than real-time
[17:44:30 CEST] <BullHorn> yes
[17:44:46 CEST] <BullHorn> i run with this: ffmpeg -i D:\Streaming\nvenc.flv -c copy -copyts F:\nvenctest.mp4
[17:45:16 CEST] <gromit> Detailed log-files for all encodes: http://akamai-progressive.irt.de/irt_reference_clips_10min/LOG.xml
[17:45:23 CEST] <BullHorn> its picking up speed, hmmn
[17:45:44 CEST] <BullHorn> 800fps now so its good
[17:46:06 CEST] <gromit> Sorry, dont know if this is zhe correct channel.just going back from vdd15 paris...see ya
[17:46:40 CEST] <BullHorn> it keeps picking up speed lol, 1200fps now
[17:46:50 CEST] <BullHorn> weird
[17:47:35 CEST] <BullHorn> done at 1500 :p
[17:50:10 CEST] <BullHorn> something is still wrong though
[17:50:29 CEST] <BullHorn> the idea of converting to .mp4 is for editing + ability to fastforward or go back to any time instantly
[17:50:44 CEST] <BullHorn> but with this version of ffmpeg the file remains as slow and unrepsonive as it was in .flv container
[17:51:37 CEST] <relaxed> BullHorn: add -movflags faststart
[17:52:21 CEST] <BullHorn> what does that do? and why did it work fine without it in the past?
[17:53:08 CEST] <relaxed> it moves the index to the beginning fo the file
[17:54:09 CEST] <BullHorn> Option movflags not found.
[17:54:12 CEST] <BullHorn> err
[17:54:49 CEST] <BtbN> How does a file "remain slow and unresponsive"?
[17:54:55 CEST] <BullHorn> its 2h long
[17:55:00 CEST] <BullHorn> i want to see what happens at 1h30min
[17:55:11 CEST] <BullHorn> i push on the scrollbar and then have to wait for 2-3 minutes for it to unfreeze
[17:55:18 CEST] <BtbN> I blame your media player.
[17:55:21 CEST] <BullHorn> when i try to view files that i converted with ffmpeg a year ago, they move instantly
[17:55:25 CEST] <BullHorn> ^
[17:56:18 CEST] <BullHorn> ill go back into older versions until i find the one where it works well
[17:56:47 CEST] <BtbN> cbbd906be6150be38dfc14b6bc67dcac8da8aea4 broke it, btw.
[17:58:24 CEST] <BullHorn> build from sep 12 works at 3000fps :o
[17:59:18 CEST] <BullHorn> but nope, still not good
[18:02:40 CEST] <BullHorn> BtbN [18:52] <@Osiris> BullHorn, flv's from obs-mp do work
[18:02:55 CEST] <BullHorn> theres 2 branches of OBS and apparently works ok with the new branch
[18:02:59 CEST] <BullHorn> but most people still use the original branch
[18:03:01 CEST] <BtbN> They're not branches.
[18:04:08 CEST] <BullHorn> potato tomato
[18:07:47 CEST] <BullHorn> what happened on april 13th
[18:07:55 CEST] <BullHorn> why the archive went from 13mb to 31mb?
[18:14:49 CEST] <BullHorn> im still going back
[18:14:52 CEST] <BullHorn> nov 2014 now
[18:15:14 CEST] <BullHorn> and its still not as far back as i was previously... i remember because when i was converting, ffmpeg created a new line for each step
[18:15:21 CEST] <BullHorn> this is still just 1 line that updates as it goes
[18:15:26 CEST] <BullHorn> ill keep going back ;/
[18:17:16 CEST] <waressearcher2> I use ffmpeg-2.3 and it creates new line when converting
[18:17:38 CEST] <waressearcher2> its quite useful, you can see how it was going at some moments in time
[18:17:47 CEST] <waressearcher2> a bit of "history log"
[18:17:54 CEST] <BullHorn> im going to try 2.5.2 now
[18:18:06 CEST] <waressearcher2> try 0.0.1-pre-alpha
[18:18:10 CEST] <BullHorn> ;/
[18:27:45 CEST] <BullHorn> im starting to suspect my problem isnt ffmpeg
[18:27:50 CEST] <BullHorn> im all the way back to 2.2.2
[18:27:57 CEST] <BullHorn> i think thats older than the one i had before
[18:27:57 CEST] <BullHorn> ;x
[18:49:59 CEST] <BtbN> BullHorn, i guess you are in the OBS irc? Could you give them this link, the flv muxer in OBS is producing invalid files. Latest master is now fixed though: https://trac.ffmpeg.org/ticket/4867#comment:2
[18:51:20 CEST] <BullHorn> thanks
[19:01:41 CEST] <luc4> Hello! When configuring ffmpeg for armv7 should I disable armv6 amd armv5 optimizations? Or should I keep those? I suppose those should be compatible with armv7.
[19:31:00 CEST] <luc4> BtbN: I just rebuilt everything again but it really seems 2.7.2 -> OK, 2.8 -> KO. I compared to the changes you linked before but I see no real change& it should be ok& I changed those macros to the new name but still that files wont play& no idea at all of what the reason may be?
[19:31:50 CEST] <BtbN> You'll need to be a bit more verbose than "that file won't play".
[19:32:07 CEST] <BtbN> The answer to that would be "you're doing something wrong"
[19:44:07 CEST] <luc4> BtbN: sorry, connection issues
[19:44:13 CEST] <luc4> BtbN: hope you got my messages.
[19:54:25 CEST] <luc4> BtbN: I also tried to reecode that file and it works properly.
[20:11:49 CEST] <xenos> Hi all, I'm having an issue using the concat demuxer to join two MP4 files together. Both files have a constant framerate but the result of the concatenate with ffmpeg has a variable framerate. Does anyone know how I can make it a constant framerate?
[20:16:11 CEST] <waressearcher2> xenos: use "-r" option ?
[20:18:00 CEST] <BullHorn> Jim- :p
[20:19:15 CEST] <xenos> waressearcher2: Same issue I'm afraid.
[20:23:30 CEST] <BullHorn> anyway thanks for the help BtbN, ill see myself out
[22:47:48 CEST] <kapusta> hi guys
[22:48:33 CEST] <kapusta> any one familiar with alphamerge around? was advised yesterday to read on it to figure my problem out but as I'm not very familiar with ffmpeg I just can't seem to nail the problem :(
[22:50:48 CEST] <durandal_1707> alphamerge-picks two inputs
[22:50:57 CEST] <kapusta> got a sequence of targa files that been split to key and fill sequences with imagemagick. i'm trying to encode both using filter_complex alphamerge and it just overlays both
[22:51:04 CEST] <kapusta> got no transparency whatsoever
[22:51:17 CEST] <kapusta> yes, got two inputs running in the command
[22:51:36 CEST] <kapusta> and when I shift one of them right and down i can see they just get overlayed
[22:51:41 CEST] <kapusta> but there's no transparency going on
[22:52:03 CEST] <kapusta> so im not sure whether im missing a parameter for the color that needs to go transparent? or maybe something else?
[22:52:37 CEST] <durandal_1707> I have hard time understanding what you need
[22:52:48 CEST] <kapusta> it could be that the split to key and fill was done incorrectly
[22:53:12 CEST] <kapusta> it probably be easier to explain via example ;)
[22:54:06 CEST] <durandal_1707> what format is key and fil?
[22:55:17 CEST] <kapusta> targa
[22:55:23 CEST] <kapusta> lemme up examples somewhere
[22:55:34 CEST] <kapusta> it might be that the alpha split was done incorrectly or something
[22:56:18 CEST] <durandal_1707> I mean pixel format
[22:59:11 CEST] <kapusta> sorry, i'm probably a little bit out of my depth here as i'm not dealing with this stuff on a regular basis
[22:59:17 CEST] <kapusta> what do you mean by pixel format?
[23:01:17 CEST] <durandal_1707> what is shown when you do 'ffmpeg -i input'?
[23:03:25 CEST] <kapusta> Stream #0:0: Video: targa, bgr24, 1920x1080, 25 tbr, 25 tbn, 25 tbc
[23:04:08 CEST] <kapusta> that was for key
[23:04:12 CEST] <kapusta> Stream #0:0: Video: targa, bgra, 1920x1080, 25 tbr, 25 tbn, 25 tbc
[23:04:15 CEST] <kapusta> and this is for fill
[23:10:37 CEST] <durandal_1707> so you want to overlay one over another using alpha components?
[23:11:11 CEST] <durandal_1707> Then use overlay filter
[23:11:39 CEST] <kapusta> tried that too
[23:11:57 CEST] <kapusta> and it just overlays inputs but gives me no transparency where it's needed
[23:12:06 CEST] <kapusta> all colours are solid
[23:12:57 CEST] <kapusta> uploaded sample key and fill tga on wetransfer
[23:13:00 CEST] <kapusta> http://we.tl/oTmMo9n3bX
[23:13:43 CEST] <kapusta> maybe by having a look somebody would catch something very obvious like these not really being suitable for alphamerge
[23:15:26 CEST] <kapusta> basically, the end result in a video would be just the rectangle object beng visible and everything else completely transparent. so what's solid white in fill would need to become transparent which i'm guessing would come from key file
[23:15:42 CEST] <durandal_1707> alphamerge takes alpha from one input and adds it to another
[23:17:39 CEST] <kapusta> i couldn't find any additional parameters for alphamerge so im assuming there is none
[23:18:07 CEST] <kapusta> and im thinking that it might be the case that the key TGA is not actually a proper alpha layer
[23:18:56 CEST] <kapusta> thats why i can't get it to work but I would love for someone who really know this stuff to reconfirm as I've been playing with ffmpeg for last 12 hours getting nowhere ;)
[23:21:22 CEST] <llogan> what is the final result supposed to look like?
[23:22:16 CEST] <kapusta> basically, if you look at the fill TGA. the rectangle object in the bottom half of the screen should be visible and everything else(solid white) should be completely transparent
[23:22:41 CEST] <kapusta> rectangle is RGB so looks proper
[23:24:06 CEST] <kapusta> i was hoping that layering it over an alpha TGA would do the trick. solid black from key would equal the transparent parts of the screen and just wipe out white out of the fill file
[23:24:26 CEST] <kapusta> buuuut, i'm faaaaaaar from being an expert on any kind of video editing ;)
[23:24:36 CEST] <durandal_1707> looks like some sort of blending
[23:25:15 CEST] <kapusta> yeah tried using blend filter and it did osmething
[23:26:36 CEST] <kapusta> but i don't think it introduced transparency
[23:27:33 CEST] <durandal_1707> Where you need transparency?
[23:27:33 CEST] <kapusta> just, well, blended both together so RGB changed colours cause the rectangle is whiteish in key but solid white went black but i don't think it was actually transparent :(
[23:27:45 CEST] <kapusta> everywhere besides the rectangle
[23:28:17 CEST] <kapusta> so while looking at the fill .tga the white portion of the screen should be comletely see through
[23:28:30 CEST] <durandal_1707> see colorkey for adding transparency
[23:28:30 CEST] <kapusta> after encoding it would be probably perfect black
[23:29:15 CEST] <llogan> alphamerge (as the input currently is) won't do what you want because i see no relevant alpha in the input. it may have alpha channel, but it appears to be unused/empty
[23:29:27 CEST] <durandal_1707> looks like some exact equatition is needed
[23:30:19 CEST] <kapusta> llogan: so my assumption about the key/fill split not being done properly is correct, yes?
[23:30:24 CEST] <llogan> here's an example alphamerge usage if it helps: http://superuser.com/a/916450/110524
[23:31:03 CEST] <llogan> the white in the mask image is actually transparent
[23:32:27 CEST] <kapusta> hmm
[23:34:19 CEST] <kapusta> thanks guys, still not sure what to do with this but i'll have a dig
[23:34:34 CEST] <llogan> probably see colorkey as mentioned
[23:34:59 CEST] <kapusta> im not sure whether the extraction of layers was done correctly
[23:35:13 CEST] <kapusta> fill is 32bit still and key looks to be 24 bit
[23:35:19 CEST] <kapusta> so its 4 and 3 channels
[23:35:34 CEST] <kapusta> where my understanding would be 3 channels for RGB and 1 channel for alpha
[23:36:09 CEST] <kapusta> and then try to merge those together to create a 4 channel image
[23:36:29 CEST] <kapusta> yes? no? ;)
[00:00:00 CEST] --- Sun Sep 20 2015


More information about the Ffmpeg-devel-irc mailing list