[Ffmpeg-devel-irc] ffmpeg.log.20190523

[burek]

[00:10:36 CEST] <Shezi> DHE, can I share a video link with you and you can check ffprobe and then do a stream with -c copy to rtmp server to find the speed issue. The video has geo location data, I am not sure its behaving because of that over https. The file is in dropbox.
[00:34:12 CEST] <another> Shezi: is your file in an mp4 container?
[00:35:07 CEST] <Shezi> Mov file but container is Video: h264 (High)
[00:35:33 CEST] <Shezi> I saw the video has Timed Metadata codec. I assume this might be causing issue. I have the video link if you want to try.
[00:36:06 CEST] <Shezi> Seems ffmpeg might not be supporting it. :|
[00:40:31 CEST] <another> hmm.. this might be a longshot, but did you move the index to the front of the file?
[00:42:11 CEST] <Shezi> Sorry which index? I am newbie ... :)
[00:47:50 CEST] <another> did you "optimise for streaming" or whatever it may be called?
[00:48:35 CEST] <another> if not, try qt-faststart <infile.mov> <outfile.mov>
[00:50:03 CEST] <Shezi> Actually I encode the video on the fly and then stream
[00:50:39 CEST] <Shezi> So I literally don't need to convert the video file at all. What is your say?
[00:52:24 CEST] <Shezi> Using qt-faststart , I get last atom in file was not a moov atom
[00:58:05 CEST] <another> seems it already has the index in front
[01:00:59 CEST] <Shezi> I am out of answers, crawled internet, nothing seems to work
[01:01:18 CEST] <Shezi> Thanks another for help. I will keep searching for more answers.
[01:08:51 CEST] <cehoyos> Shezi: What was your original question?
[01:13:01 CEST] <Shezi> Actually I have a video file mov with h264 container. The video was shot with iPhone . It has Timed Metadata codec . I can stream the same file from my local machine to an rtmp server. But If I put the same file on http server and try to stream it, then it works but the speed goes to 0.28x and less.
[01:13:47 CEST] <Shezi> ffmpeg -re -I file_in_my_machine -c copy rtmp://some_server.    <--- This works ok
[01:14:26 CEST] <Shezi> If I do ffmpeg -re -I http://somewebserver/file_in_my_machine -c copy rtmp://some_server <---- this does not work
[01:15:21 CEST] <Shezi> File is same. On my local machine it gives me 1x speed when streaming. When trying from http server, the speed reduces to 0.28x .
[01:15:27 CEST] <cehoyos> Please paste both console outputs (complete, uncut) somewhere
[01:15:33 CEST] <Shezi> Tried encoding it, no avail.
[01:15:36 CEST] <Shezi> ok
[01:18:49 CEST] <Shezi> Can I paste it here or to your private chat?
[01:18:54 CEST] <Shezi> I mean the link
[01:19:02 CEST] <Shezi> Pastebin link.
[01:20:20 CEST] <another> why are you using -re on content via http?
[01:21:45 CEST] <Shezi> Isn't it used for streaming?
[01:22:30 CEST] <another> it reads input at realtime to simulate streaming
[01:22:36 CEST] <another> see https://ffmpeg.org/ffmpeg-all.html
[01:23:30 CEST] <Shezi> https://pastebin.com/Yi99WH1U
[01:23:38 CEST] <Shezi> https://pastebin.com/AgQkB5pW
[01:23:52 CEST] <Shezi> Same file, one ran locally and other from https server
[01:24:14 CEST] <Shezi> Ok another , I will take a look.
[01:24:22 CEST] <cehoyos> Is the issue reproducible with current FFmpeg git head?
[01:24:36 CEST] <Shezi> But that's what I want to create a realtime effect.
[01:24:58 CEST] <Shezi> cehoyos, yes I tried latest static build from johnvansickle
[01:25:16 CEST] <Shezi> git: ge45e6005ce built on 20190521
[01:25:16 CEST] <cehoyos> No, the version you tested is quite old
[01:25:27 CEST] <Shezi> I tested git: ge45e6005ce built on 20190521 this as well
[01:25:46 CEST] <cehoyos> 13 months
[01:26:46 CEST] <Shezi> I tested this latest build as well
[01:27:18 CEST] <Shezi> I can share dropbox link to that file. You can also try if you want to.
[01:27:47 CEST] <Shezi> Seems very confusing. I tried it directly from dropbox , I thought it might be my web server. But same effect.
[01:29:36 CEST] <another> Shezi: "Should not be used with actual grab devices or live input streams (where it can cause packet loss)."
[01:30:43 CEST] <Shezi> I am not using grab devices.
[01:30:54 CEST] <Shezi> :)
[01:32:48 CEST] <another> ah, i misread
[01:33:02 CEST] <another> but try without -re anyway
[01:33:41 CEST] <Shezi> Same effect, just tried it
[01:34:24 CEST] <Shezi> Now the fun part is, if I convert this video with -c copy and put it in web server it works with https.
[01:34:47 CEST] <another> covert how?
[01:34:49 CEST] <Shezi> The file will have exact same bitrate, same fps and same duration and size.
[01:35:00 CEST] <Shezi> Ffmpeg -I input -c copy output
[01:35:21 CEST] <another> are you putting it in a different container?
[01:35:28 CEST] <Shezi> no
[01:35:34 CEST] <Shezi> Same one
[01:35:41 CEST] <Shezi> -c copy puts it in same container
[01:35:49 CEST] <another> no
[01:35:51 CEST] <th0rne> In a thread about using atomicparsley to set video covers, random guy on Web says, "Just don't remux the file with ffmpeg afterwards". Why?
[01:36:23 CEST] <another> -c copy tells ffmpeg to copy streams
[01:36:50 CEST] <Shezi> but I checked the container was same, although profile changed.
[01:37:13 CEST] <furq> th0rne: the top line of the answer you're quoting from links an unresolved ffmpeg bug report
[01:37:14 CEST] <another> the container is deduced from the file extension, unless explicitly specified
[01:37:16 CEST] <furq> that's why
[01:37:25 CEST] <Shezi> Yes that I know -c copy does copy the streams. You can tell ffmpeg to copy only audio or video or both.
[01:38:28 CEST] <furq> it was unresolved when the answer was added anyway
[01:38:34 CEST] <furq> maybe it works now
[01:40:06 CEST] <cehoyos> If removing re doesn't help, I suggest you test current FFmpeg git head and post both command lines including complete, uncut console output on the user mailing list.
[01:41:11 CEST] <Shezi> Yes removing re didn't help. Instead of johnvansickle, I will compile from git head and will share result. I have to sleep now. Its midnight 2:41. But I will share results tomorrow.
[01:41:26 CEST] <th0rne> furq: Ok, I see, thank you!
[01:43:12 CEST] <Shezi> another, thanks for your help. Will catch you guys tomorrow with more tests on latest ffmpeg
[01:43:20 CEST] <Shezi> Ciao :)
[03:30:41 CEST] <riff-IRC> Hey all, need some help with preset files.
[03:31:17 CEST] <riff-IRC> I want to create a preset file to make FFmpeg use my GPU.
[03:43:26 CEST] <CounterPillow> oh lord
[03:44:05 CEST] <DHE> ffmpeg has preset files?
[03:44:29 CEST] <riff-IRC> http://ffmpeg.org/ffmpeg-all.html#Preset-files
[03:44:50 CEST] <riff-IRC> Dealing with the cluster that is Kdenlive and GPU rendering.
[03:47:10 CEST] <riff-IRC> I want to specify the following encoder option: vaapi_device=/dev/dri/renderD128
[03:47:41 CEST] <riff-IRC> How would I make a preset file that would pass that flag on to FFmpeg?
[07:42:14 CEST] <Gigabitten> So I've got a bunch of images and I'm trying to make them into a slideshow. They're all the same size. It's very simple. But I can't get it to be longer than about half an hour for some reason.
[07:43:01 CEST] <Gigabitten> I'm doing `ffmpeg -framerate 0.03583333 -pattern_type glob -y -i '*.jpg' ../2.mp4
[07:43:05 CEST] <Gigabitten> `
[07:43:31 CEST] <Gigabitten> I've tried using a different file format, I've tried cutting some images out around where it was failing, no dice.
[10:42:44 CEST] <sine0> I want to process with ffmpeg. I want to be able to process just single frame somewhere in the video so I can test out settings and then when happy with frame run it across the whole conversion, does that make sense
[10:46:28 CEST] <sine0> ok i got it  ffmpeg -ss 00:27 -i MVI_5230.MOV -vframes 1 -q 2 sample.jpg
[11:01:26 CEST] <sine0> mp=eq2 is the rgb right ?
[11:06:58 CEST] <sine0> ok im stuck how do I edit RGB values
[11:13:24 CEST] <sine0> I am not sure how to interpret bug trac or suggestions but I guess this has not been included? https://trac.ffmpeg.org/ticket/4858
[13:42:45 CEST] <jpiat> Hi, i'am using libavformat/libavcodec to generate a RTMP stream from H264 NALUs and AAC encoded audio (in a flv container). When doing so, only ffplay gets to play the stream with audio and video. Other players would just play the audio. If is use ffmpeg to copy the input stream to a new rtmp url (without re-encoding) it works fine for all players. Any idea of what the ffmpeg tool does that i don't in the original stream ?
[13:58:41 CEST] <furq> jpiat: at a guess, https://www.ffmpeg.org/ffmpeg-bitstream-filters.html#h264_005fmp4toannexb and/or https://www.ffmpeg.org/ffmpeg-bitstream-filters.html#aac_005fadtstoasc
[14:14:07 CEST] <jpiat> furq: i'am converting my annexb NALU to AVCC already otherwise the FLV container would just not accepts those when muxing.
[15:01:57 CEST] <upgreydd> Hello all. I have a question. I have some file with `ani` externsion. In this file I have 42bytes of custom header? and after that h264 header. File plays fine in any player but what format is that? I found some data in this header like file size in bytes and number of frames, but I'm not sure what kind of format is this. Any advice please?
[15:17:35 CEST] <cehoyos> upgreydd: Please provide a sample file
[15:20:57 CEST] <upgreydd> cehoyos: any special place to upload or doesn't matter?
[15:21:15 CEST] <cehoyos> Must not need login, no encryption please
[15:23:46 CEST] <furq> upgreydd: if it plays back in something that uses ffmpeg libs (e.g. mpv) then just run ffprobe on it
[15:25:35 CEST] <sine0> I have some raw video footage, is there anyway of me finding out the information about it
[15:25:57 CEST] <sine0> I have to tell ffmpeg what it is using pix_fmt
[15:26:42 CEST] <upgreydd> furq: ffprobe is not showing additional 41bytes specification
[15:27:02 CEST] <furq> it should tell you what container it thinks it is
[15:29:14 CEST] <upgreydd> furq: cehoyos: https://filebin.net/d1975zaw6liyhqim here is example file it's ~370kb
[15:30:40 CEST] <upgreydd> cehoyos: in offset 4-7 there's 16bit Little Endian - number of frames, 8-11 file size in bytes. No idea what's rest of 42 bytes
[15:30:56 CEST] <upgreydd> cehoyos: from 43 byte h264 begins
[15:49:54 CEST] <upgreydd> cehoyos: any idea? :D
[15:52:17 CEST] <upgreydd> this file comes from qnx/blackberry system
[15:53:51 CEST] <cehoyos> I have never seen such files, it appears to me that they work well with FFmpeg.
[15:54:02 CEST] <cehoyos> I have to go but I will see further comments from you later
[16:19:18 CEST] <sine0> im using 3.1.4 am I elite
[16:24:06 CEST] <sine0> does interpolate blend 2 frames to create additional frame?
[16:43:24 CEST] <cehoyos> sine0: Do you mean minterpolate?
[16:44:11 CEST] <sine0> yes
[16:44:39 CEST] <sine0> I have the latest build after using an old one and I have used minterpolate on a 12.5 fps to make it 25fps
[16:45:14 CEST] <cehoyos> If you use minterpolate to increase framerate, motion compensation algorithms will be used to create additional frames
[16:45:15 CEST] <sine0> if i save with x264 the video is just grey data mess, with green bursts. I have tried avi and that works although the setting was low qual
[16:45:42 CEST] <cehoyos> upgreydd: Not sure what you mean with "from 43 byte h264 begins" - the header has 588 bytes afaict
[16:45:56 CEST] <sine0> wow the minterpolate, I must say, I am impressed!
[16:46:49 CEST] <upgreydd> cehoyos: in offset 42 (43 byte) you have 00 00 00 01 67 which is h264 format header
[16:47:46 CEST] <upgreydd> cehoyos: I have few `ani` files and: offsets: 0-3 - static 4-7 - number of frames 8-11 - file size i don't know what's next :/
[16:50:10 CEST] <upgreydd> cehoyos: it can be related to some specific blackberry/qnx encoding, not sure what's that :/
[16:52:09 CEST] <upgreydd> cehoyos: or it can be related to qt
[16:52:30 CEST] <upgreydd> qt - not a quicktime, the linux framework qt
[16:52:32 CEST] <cehoyos> Does not look like qt
[16:52:45 CEST] <cehoyos> I don't know much about the qt framework
[16:52:54 CEST] <cehoyos> I would have expected the resolution but I don't see it
[16:53:31 CEST] <upgreydd> no idea what's this header related to, tried to decipher whole header to prepare my animation but it doesn't work - I have black screen :(
[16:54:39 CEST] <cehoyos> Why do you want to decipher it: It does not seem to contain useful information, no?
[16:55:00 CEST] <cehoyos> You are of course right about the 42 byte header
[16:56:32 CEST] <upgreydd> cehoyos: I converted my video with this params: https://0bin.net/paste/7c2dEY8XcoM5XSvo#XsW5rGjbJmK3RM0ppTU4Qb7lQ3eFi+gpSI34pMISa0p and it looks same in ffprobe like the ani file, but it wouldn't play
[16:57:18 CEST] <cehoyos> Does not look like an FFmpeg command line including complete, uncut console outpu;-/
[16:57:31 CEST] <cehoyos> Do you want to create such files? Why?
[16:58:24 CEST] <upgreydd> cehoyos: customizing some hardware. cehoyos: does libx264 differs with other h264 codecs? I'm newbie in video encoding so meybe I should use something different? I've found that qnx is using ipp_h264 decoder
[16:59:31 CEST] <cehoyos> I suggest you ask whoever wants the files, it may or may not be possible to fulfill the spec with FFmpeg / x264 and some header magic
[17:00:04 CEST] <upgreydd> cehoyos: https://0bin.net/paste/H6oGYfPR+mruLTY5#gGHdKkVzgJDl4EUVLyUYaQZnfcXtF5g2vCIxz27ue3Q that's command which I've used
[17:02:43 CEST] <cehoyos> Does not look like an FFmpeg command line including complete, uncut console output;-(
[17:03:05 CEST] <upgreydd> cehoyos: does libx264 encoding differs with h264 ?
[17:03:43 CEST] <cehoyos> x264 is an encoder implementation of the h264 standard, to the best of my knowledge, it is the most complete implementation apart from PAFF
[17:03:51 CEST] <cehoyos> and "best"
[17:06:10 CEST] <upgreydd> cehoyos: ok and what's ipp_h264? I found some info about intel decoder but I'm not sure what's that and it can be a reason why movie is black/not playing
[17:06:36 CEST] <JEEB> check the non-playing video with ffprobe -v verbose ?
[17:06:52 CEST] <JEEB> and post the ffprobe result on padtebin and link here
[17:07:00 CEST] <cehoyos> It's Intel's implementation, see very old Russian quality paper for a comparison
[17:07:22 CEST] <cehoyos> Please post the ffmpeg command that creates the file, much more information...
[19:14:08 CEST] <Shezi> cehoyos: I have made new tests with ffmpeg master branch.
[19:14:21 CEST] <Shezi> https://pastebin.com/gRGRGd8i   <--- using url
[19:14:47 CEST] <Shezi> https://pastebin.com/cG4fzdp4 <--- same file from local file system
[19:15:30 CEST] <Shezi> commit id : 82ccb9cba9a7bd530ba2070c90f03d0fde9af1bd
[19:16:00 CEST] <Shezi> There is one commmit more on top of it. 8f6e65183354d1d402ae80c71cba2759fe152018
[19:16:09 CEST] <Shezi> Same result unfortunately.
[19:33:36 CEST] <Shezi> I also found that ffmpeg is try again and again the connection which is causing the speed to slow down.
[19:57:05 CEST] <nadermx> Hi all, I asked on superuser the issue I'm having, basically ffmpeg is merging the audio and video but instead of adding all the audio parts of the video are silent and then the auido returns https://superuser.com/questions/1439735/merging-audio-and-video-with-ffmpeg-causes-audio-to-miss-parts-or-be-silent
[20:09:54 CEST] <xochilpili> hello everyone, im testing an m3u8 playlist using ffprobe but got a loop "last message repeated 246 times" < keep increasing, is there a way to avoid this ?
[20:10:48 CEST] <DHE> look at the error directly above it, and resolve the problem
[20:12:25 CEST] <xochilpili> Failed to open segment 388311 of playlist 2 < that's the error: how to solve the problem in ffprobe ?
[20:13:27 CEST] <riff-IRC> Seems like a problem with the playlist- try making a new playlist and playing it
[20:13:28 CEST] <DHE> files listed in the playlist are missing or otherwise unreadable. fixing that is on you I think
[20:13:49 CEST] <DHE> assuming they're on a web server, its logs may help
[20:13:51 CEST] <riff-IRC> BTW, what brand of hard drive are you storing the playlist on?
[20:14:03 CEST] <xochilpili> that's not my playlist, i cant fix it
[20:14:12 CEST] <riff-IRC> Oh
[20:16:11 CEST] <xochilpili> im making a bash script to check sources in m3u8 playlists, and recreate without those which ffprobe fail, but how can i avoid that looping in ffprobe ?
[20:44:24 CEST] <Shezi> It seems ffmpeg has issues with iPhone video files streaming from https url.
[23:03:22 CEST] <cehoyos> Shezi: You can run your command (the one that you inexplicably decided not to show us in the console output) with -loglevel verbose to verify that reconnects because of necessary seeking for a badly interleaved input file is the reason for the slowdown.
[23:22:11 CEST] <Shezi> cehoyos, yes it reconnects
[23:22:22 CEST] <Shezi> Can we some how fix this?
[23:22:57 CEST] <cehoyos> Please show the console output.
[23:28:04 CEST] <steve___> cehoyos: you have the patience of a saint
[23:29:18 CEST] <JEEB> with mov/mp4 the demuxer just seeks a lot so you need some sort of cache if you want to make it quicker
[23:29:31 CEST] <JEEB> (also some files are funkily interleaved)
[23:30:29 CEST] <cehoyos> ... with -loglevel verbose
[23:41:28 CEST] <Shezi> Ok sure
[23:49:26 CEST] <Mysterytrain> I have a question. why doesn't ffmpeg follow stdin/stdout conventions? You can't go ffmpeg-i somefile.mkv > report.txt. or similar
[23:49:49 CEST] <Mysterytrain> err, ffprobe -i I mean
[23:50:02 CEST] <BtbN> Because you could output a media file to stdout
[23:50:08 CEST] <BtbN> all logging goes to stderr
[23:50:48 CEST] <Mysterytrain> ah, I seee
[23:50:58 CEST] <Mysterytrain> well at least there is a -report option.
[23:51:13 CEST] <BtbN> There is also 2>...
[23:55:26 CEST] <Shezi> cehoyos, here you go. https://pastebin.com/FxVTSc7M
[23:55:46 CEST] <cehoyos> Statistics: 915197 bytes read, 27 seeks
[23:55:59 CEST] <cehoyos> That doesn't sound like the reason it doesn't work well
[23:56:37 CEST] <cehoyos> Otoh, this was for one second, I suggest to let it run longer
[23:56:51 CEST] <cehoyos> Do it locally (not over http), to see the output quicker
[23:57:07 CEST] <Shezi> Locally It works fine
[23:57:09 CEST] <cehoyos> (Seeking at the start is not unusual)
[23:57:35 CEST] <cehoyos> That's not the point: If there is an issue with interleaving, you can also see it locally.
[23:58:09 CEST] <Shezi> ok
[23:58:33 CEST] <Shezi> So do you need to see the locally tested file output?
[23:59:23 CEST] <cehoyos> It doesn't matter: Both outputs would show the issue (if interleaving is the issue)
[23:59:32 CEST] <cehoyos> But locally it runs faster iiuc
[00:00:00 CEST] --- Fri May 24 2019