[Ffmpeg-devel-irc] ffmpeg.log.20180807

[burek]

[00:10:19 CEST] <ArsenArsen> there is a transcode example but it does a load more than just transcode (like do filters)
[00:16:03 CEST] <DHE> they're examples. prune them, mix and match them
[02:13:32 CEST] <pi-> https://paste.pound-python.org/show/rn84VHFtaoubs1fh12cA/
[02:13:53 CEST] <pi-> ^ can anyone see a cleaner way of doing this?
[02:15:58 CEST] <pi-> https://paste.pound-python.org/show/Wwd5oMqY3r45kuKtv20P/
[02:16:07 CEST] <pi-> ^ this is the complete function
[02:16:23 CEST] <pi-> What is the developer doing with args_ensure_speed?
[06:21:07 CEST] <johnnny22> Hi, I'm trying to seek at 08:48:46 using -ss on the input, but then I get a info msg saying "could not seek to position 113584.864" . Why is it trying that time instead of 31726 ?
[06:23:48 CEST] <johnnny22> the input is an HLS stream that is 'ended'
[06:29:03 CEST] <johnnny22> oh humm
[06:32:29 CEST] <johnnny22> I'm actually not totally grasping the situation, but I'm guessing it partially has to do with the fact that the streams props are as follow: Duration: 03:59:52.00, start: 81858.864056
[06:32:47 CEST] <johnnny22> I basically want to seek at a given timestamp :o/
[06:34:01 CEST] <Cracki> you want to seek to EIGHT hours into a stream?
[06:34:05 CEST] <Cracki> are you sure it's that long?
[06:34:37 CEST] <Cracki> why do you want to seek to a time that isn't there?
[06:34:45 CEST] <Cracki> no, really, I'm asking you WHY
[06:34:56 CEST] <Cracki> not a rhetorical question
[06:34:59 CEST] <johnnny22> well, to be honest, i want to seek at 22:48:46.13  which is a timestamp in the stream.
[06:35:22 CEST] <johnnny22> humm
[06:35:25 CEST] <Cracki> how long is your stream
[06:35:40 CEST] <johnnny22> according to ffprobe just under 4hours
[06:35:46 CEST] <Cracki> and, is it?
[06:35:57 CEST] <johnnny22> ya
[06:36:05 CEST] <Cracki> so?
[06:36:12 CEST] <Cracki> does what you want make sense?
[06:36:47 CEST] <johnnny22> but the 1st timestamp is around 22:43:00 or so
[06:37:15 CEST] <Cracki> what are you talking about
[06:37:40 CEST] <johnnny22> the timestamps in the mpegts, i guess the pts/dts values
[06:37:48 CEST] <Cracki> where do these "timestamps" come from
[06:37:54 CEST] <Cracki> and why do you think your file is that long?
[06:38:23 CEST] <Cracki> what does "first" timestamp even mean?
[06:38:25 CEST] <johnnny22> the stream is 4h long (3h58m)  it's start time is : "start: 81858.86405" according to ffprobe.
[06:38:34 CEST] <Cracki> why didn't you say so?
[06:38:43 CEST] <johnnny22> sorry if i wasn't clear
[06:39:18 CEST] <johnnny22> I did specify the stream props above, but eh :)
[06:39:32 CEST] <johnnny22> i didn't state everything clearly at the beginning
[06:42:10 CEST] <johnnny22> so, if I recap: My stream is 3h58m long and starts at 81858.86405 according to ffprobe. This means the stream starts around time 22:43:00 or so. I'd like to seek the input using "-ss 22:48:46.13" on the input, but when I try this using -v info, i get an error message saying "could not seek to position 163984.994" .
[06:42:29 CEST] <johnnny22> I'm using -copyts too.
[06:42:56 CEST] <johnnny22> If I use : "-ss 22:48:46.13" on the output, it actually works.
[06:43:17 CEST] <johnnny22> But then, my outputted timestamps start at 0 and the copyts seems to have no effect.
[06:49:50 CEST] <johnnny22> to be honest, i don't mind using it on the output, but I'd need the timestamps to be right too
[06:52:35 CEST] <johnnny22> i'll play around some more with different output formats to see clearer.
[06:59:19 CEST] <johnnny22> i'm really baffled by this 163984.994 value it comes up with
[07:02:07 CEST] <johnnny22> i guess on the input, the ss position can only be a position from beginning, while on the output, it can apply to the timestamps too.
[07:04:53 CEST] <johnnny22> or at least with i copyts of the input
[08:04:54 CEST] <freedrull> i'm overlaying a gif on a video, it works but the framerate of the gif seems lower in the output ? is there any way i can address this? https://gist.github.com/mcfiredrill/6b50981e030d0230b9c1d8d9cd1559ea
[08:21:33 CEST] <fred__tv> Hi, very new to ffmpeg : is it possible to play an ip camera rtsp stream into a device framebuffer ?   any documentation ?
[14:05:38 CEST] <Nacht> I'm trying to use the basic rotated text overlay using an alphamerge filter. However, I notice that due to the text being on a black background, the eventual text has somewhat black edges. Any way to remedy this ? I tried changing the background color, but then alphamerge doesn't work anymore.
[14:06:30 CEST] <Nacht> This is the command I use: https://pastebin.com/pBCK2BBp
[14:40:52 CEST] <tomtomb> Hi All, i'm trying to concat a bunch of files, and getting the error "Application provided invalid, non monotonically increasing dts to muxer in stream XX" a lot of times and it seems the concat is stuck... is there any way to fix this error?
[14:42:13 CEST] <BtbN> it's not an error.
[14:42:38 CEST] <BtbN> If you concat a bunch of unrelated files, of course the timestamps don't match up
[14:45:58 CEST] <tomtomb> BtbN, oh! gotcha, thanks!, i'm also getting the following messages, "edit unit sync lost on stream 0, jumping from 4459 to 4458", "cannot find current edit unit for stream 0, invalid index?", do those make sense too?
[15:36:36 CEST] <tomtomb> BtbN, is there any way to "unsubscribe" from those messages?
[17:26:23 CEST] <termos> Is there a way to get x264 to share encoding data across renditions? I have the same video transcoded several times, I figure I can re-use the motion vector and macro block-divison steps for example
[17:30:21 CEST] <DHE> termos: I believe that's a thing in x265 (at the API level) but not x264
[17:30:34 CEST] <termos> I see, thanks
[17:30:42 CEST] <DHE> not sure how shared MVs would work if you're rescaling the picture though
[17:33:44 CEST] <termos> was hoping I could scale those, the scale transform should not change directions afaik
[18:07:17 CEST] <acetone> Hi all!, I am encoding raw images in C to mjpeg avi, I figured out conversion and stuff, but I am getting bottom half of the picture purple
[18:07:22 CEST] <acetone> like this: https://imgur.com/a/6rZWEBP
[18:07:35 CEST] <acetone> do you know, what the issue may be?
[18:08:37 CEST] <Mavrik> acetone: that means that your picture buffers aren't the right size
[18:08:46 CEST] <Mavrik> You're only filling out half of the picture :)
[18:15:31 CEST] <acetone> thanks, for the yuv420p it should be x*y*4 bytes?
[18:16:39 CEST] <atomnuker> no, stride[0]*height + stride[1]*(height >> 1) + stride[2]*(height >> 1)
[18:20:18 CEST] <acetone> atomnuker: thanks, well AVframe seems to be set up correctly. But I think, that I am approaching this from wrong angle
[18:20:53 CEST] <acetone> if the image is not skewed, but cut in half
[18:21:19 CEST] <Mavrik> It just looks like you're not filling out the last half of luma (and perhaps other two channels)
[18:21:29 CEST] <Mavrik> If you'd have strides wrong, you'd get diagonal stripes
[18:22:05 CEST] <iive> unless he uses double stride, then he gets shrunk image :D
[18:22:17 CEST] <iive> but it looks proportional.
[18:24:41 CEST] <acetone> yeah proportions are OK, but now I am looking at conversion and the frame I am filling has linesize[0] = 960 linesize[1]=480 linesize[2]=480
[18:25:04 CEST] <acetone> and it should be half of 1920x1080
[18:25:43 CEST] <acetone> that seems to be OK
[18:30:52 CEST] <acetone> for some reason I am flipping the frame and doing some color conversion - I will try to disable that
[18:32:16 CEST] <acetone> also scale context is from 1920x1080 to 960x544 instead of 960x540, but afaik it should be. colors from AV_PIX_FMT_BGR32 to AV_PIX_FMT_YUVJ420P
[19:26:06 CEST] <yagiza> Hello!
[19:27:21 CEST] <durandal_1707> Hello!
[19:28:04 CEST] <yagiza> Is this the correct place to ask questions about using FFMpeg/LIBAV API?
[19:28:51 CEST] <Mavrik> mhm
[19:31:03 CEST] <yagiza> Mavrik, was that "yes" or "no"?
[19:33:58 CEST] <yagiza> It seems noone's around to answer.
[19:34:03 CEST] <yagiza> Anyway.
[19:34:09 CEST] <yagiza> I'll try.
[19:34:54 CEST] <yagiza> Do any1 around here have experience with RTP stream demuxing?
[19:35:20 CEST] <yagiza> If no, maybe someone can suggest me a correct place to ask?
[19:35:46 CEST] <durandal_1707> do you have some kind of bug to report?
[19:36:03 CEST] <yagiza> I have a problem and I can't find a place to get help.
[19:36:30 CEST] <yagiza> I'm not sure if it's a but in the liubrary, or just I'm doiung something wrong.
[19:36:49 CEST] <durandal_1707> what kind of problem?
[19:36:54 CEST] <yagiza> There is a feature, which I cant get working.
[19:37:14 CEST] <yagiza> And there is no working example of using it .
[19:37:59 CEST] <yagiza> The feature is "sdp_flags=custom_io" options of rtsp demuxer.
[19:39:21 CEST] <yagiza> When I specify the option, giving an SDP file, the option is accepted and SDP processed correctly, but nothig happens.
[19:40:35 CEST] <yagiza> It reads SDP data from my custom I/O procedures, but then it just listens IP and port specified in SDP instead of reading RTP data from my custom I/O.
[19:40:57 CEST] <yagiza> Just like no option specified.
[19:42:05 CEST] <yagiza> And I don't know where to ask about it.
[20:14:27 CEST] <acetone> Just for the record - I had srcSliceH in sws_scale() set to target instead of source... RTFD, or the joy of glueing the code.
[20:14:40 CEST] <acetone> Anyway, thanks everyone :)
[20:15:30 CEST] <acetone> I guess tomorrow I will be here trying to do the same process with png codec...
[22:09:33 CEST] <wingrime> Hi all!
[22:09:58 CEST] <wingrime> I have some problems with rtmp streaming
[22:10:11 CEST] <wingrime> constant framedrops
[00:00:00 CEST] --- Wed Aug  8 2018