[Ffmpeg-devel-irc] ffmpeg.log.20180322

burek burek021 at gmail.com
Fri Mar 23 03:05:02 EET 2018


[00:03:01 CET] <JEEB> keyframes just mean that that specific frame can be decoded given the decoder initializer data (which might or might not be required depending on the A/V format)
[00:03:41 CET] <JEEB> random access point is something that specifies that that as you start decoding from it, all frames following it in presentation order can be successfully decoded
[00:08:37 CET] <Guest66109> so, if I get that frame, that is a random access point, I can cut down the video at that point witout borked pictures. Can I find these frames with ffmpeg?
[00:09:48 CET] <JEEB> with something like ffprobe, yes
[00:10:07 CET] <JEEB> ffprobe -of json -show_streams -show_packets FILE
[00:10:17 CET] <JEEB> gives you JSON you can parse with some script
[00:20:17 CET] <Guest66109> but what is the flag I should be looking for in ffprobe output? So far I have not seen anything there that tells me its different frame from the rest
[00:33:32 CET] <Quenz> I'm trying to convert a gif to an mp4, using "ffmpeg -i input.gif -c:v libx264 -r 30 -pix_fmt yuv420p output.mp4 -vf scale=1200:-2", but got the error "Error while opening encoder for output stream #0:0 - maybe incorrect parameters such as bit_rate, rate, width or height"
[00:36:02 CET] <Quenz> Here's the full terminal log for the command https://pastebin.com/y2pq8xft
[00:36:33 CET] <furq> Quenz: -vf goes before the output filename
[00:38:03 CET] <Quenz> Thanks, that fixed it furq
[01:20:14 CET] <ZexaronS> Hello
[01:21:17 CET] <ZexaronS> I'm wondering if I could make some kind of a script that would cut all audio files timecode, leaving the encoding alone, but there's several points in a single file, 3 or 4
[01:21:26 CET] <ZexaronS> batch script
[01:38:40 CET] <nurupo> durandal_1707: `ffmpeg -i original.mp4 -af silenceremove=0:0:0:-1:20:-50dB -b:a 160k -cutoff 24k silenceremove4.mp4`
[01:39:56 CET] <nurupo> it would be great if i could cut out the silent segments from my aac stream without having to re-encode the audio, but this is good enough
[04:37:46 CET] <Renich> I have tried animating text here but can't: http://dpaste.com/0B0CTHY
[04:37:59 CET] <Renich> When I add any animation, it doesn't happen. Is it because the input is a single file?
[04:43:38 CET] <furq> Renich: add -loop 1 before -i ${image_path}
[04:43:46 CET] <furq> otherwise you'll just end up with one frame in the output
[04:43:53 CET] <Renich> OK
[04:44:03 CET] <furq> also you shouldn't need -strict -2 any more unless your ffmpeg is really old
[04:44:11 CET] <furq> and if it is that old you should upgrade
[04:44:19 CET] <Renich> furq: thanks. Will remove it as well
[04:45:56 CET] <Renich> furq: I've added -loop 1 and now it's rendering it far past the duration of the song
[04:46:02 CET] <Renich> 3 minutes so far
[04:46:03 CET] <furq> oh right yeah
[04:46:04 CET] <furq> add -shortest
[04:46:14 CET] <furq> as an output option
[04:46:24 CET] <Renich> OK
[04:47:24 CET] <Renich> furq: OK, that worked. Let me try some animation now.
[04:48:15 CET] <Renich> yeah! it works! you rock!
[04:55:45 CET] <Johnjay> TIL how to use the -shortest option
[04:55:49 CET] <Johnjay> furq you truly do rock
[05:26:46 CET] <nurupo> how can i prevent ffmpeg from printing the progress lines?
[05:27:30 CET] <nurupo> i have cron'ed it and it fills up my logs with lines like "size=    1344kB time=00:01:08.00 bitrate= 161.9kbits/s speed=1.06x"
[05:28:57 CET] <nurupo> i want to see the stats and how much it took ffmpeg in total to re-encode the file, but not the intermediate progress
[05:37:23 CET] <furq> nurupo: -nostats
[05:37:31 CET] <furq> although it sounds like you just need to make your terminal wider
[05:39:10 CET] <kurufu> its being logged to a file, so i dont think thats his issue.
[05:39:32 CET] <furq> well yeah that'll do it as well
[06:59:17 CET] <bodqhrohro> Is random common for all expressions? I'm getting an identical noise here, probably because of same seed: nullsrc=s=80x360, geq=lum_expr=128:cb_expr=128+tan(random(1)*1.5)*9:cr_expr=128[blue_noise];
[06:59:17 CET] <bodqhrohro> nullsrc=s=80x360, geq=lum_expr=128:cr_expr=128+tan(random(2)*1.5)*9:cb_expr=128[red_noise];
[06:59:17 CET] <bodqhrohro>  Can I init it with different seeds?
[07:04:09 CET] <nurupo> ><nurupo> i want to see the stats
[07:04:12 CET] <nurupo> ><furq> nurupo: -nostats
[07:04:21 CET] <nurupo> huh
[07:05:24 CET] <furq> if you just want to see the last statusline then i don't think that's supported if you're writing to a logfile
[07:05:42 CET] <furq> i figured you meant the x264 stats etc
[07:34:57 CET] <skywalker> i'm a newbie to ffmpeg, how can I add stepping frame by frame backwards to ffplay.c?
[07:35:13 CET] <skywalker> i'm confused, don't know where to start
[07:43:50 CET] <wfbarksdale> you've much to learn, young skywalker
[07:45:01 CET] <skywalker> @wfbarksdale, what's your recommendations? can I finish the stepping-by-frame-backwards in 8 days?
[07:45:19 CET] <skywalker> @wfbarksdale, where should I start?
[09:03:14 CET] <hellyeah> i will download ffmpeg what is the difference between static and shared?
[09:03:48 CET] <pmjdebru1jn> whether the libraries are built into the binary
[09:03:57 CET] <pmjdebru1jn> as in static vs dynamic linking
[09:04:12 CET] <hellyeah> so i can download static
[10:11:21 CET] <azarus> Ugh. This ticket bit me today: https://trac.ffmpeg.org/ticket/5718
[10:11:44 CET] <azarus> open since 20 months...
[10:27:24 CET] <dealer> [NULL @ 0x298e1a0] Unable to find a suitable output format for 'binary' binary: Invalid argument
[10:27:31 CET] <dealer> what the heck is going on
[10:27:33 CET] <dealer> any help
[10:34:30 CET] <dealer> ?????
[10:35:20 CET] <durandal_1707> dealer: post full console output
[10:35:27 CET] <durandal_1707> via pastebin
[10:35:44 CET] <dealer> ok
[10:37:57 CET] <dealer> https://pastebin.com/YxTUnDEu
[10:40:17 CET] <dealer> any help
[10:40:48 CET] <durandal_1707> dealer: no help for you, where is full command output?
[10:41:31 CET] <dealer> this is full brother when i type ffmpeg binary
[10:41:31 CET] <durandal_1707> i need full ffmpeg command not just output, input too
[10:41:48 CET] <durandal_1707> you mean you just type: ffmpeg
[10:41:49 CET] <durandal_1707> ?
[10:42:27 CET] <durandal_1707> also that is very old version, from 2016
[10:44:14 CET] <dealer> no durandal_1707  i typed ffmpeg binary
[10:44:26 CET] <dealer> basically binary files are missing
[10:44:37 CET] <dealer> durandal_1707:  can u guide me where is the new version from where i can install
[10:44:48 CET] <dealer> https://trac.ffmpeg.org/wiki/CompilationGuide/Centos
[10:44:51 CET] <dealer> i followed this
[10:48:36 CET] <kwizart> dealer, you can get pre-built binary for centos using rpmfusion.org for EL
[10:49:12 CET] <dealer> kwizart:  how ? I am not that techy
[10:49:33 CET] <durandal_1707> dealer: use git to get latest source if you want master ffmpeg version
[10:50:01 CET] <kwizart> dealer, see https://rpmfusion.org/Configuration , then yum install ffmpeg (you will get 2.8.14), but you can also build from source
[10:51:42 CET] <dealer> Package ffmpeg-2.6.8-3.el7.nux.x86_64 already installed and latest version Nothing to do
[10:52:36 CET] <kwizart> dealer, nux is a unmaintained repo, please avoid to use that, use rpmfusion.org instead
[10:54:14 CET] <dealer> i should remove it and start from zero
[10:54:52 CET] <kwizart> dealer, probably
[10:55:03 CET] <dealer> can you guide me
[11:32:50 CET] <TarquinWJ> hi, I am trying to get ffmpeg to create M4A AAC-LC *mono* with a single audio channel, but it doesn't seem to be working. Instead, it always creates two identical mono channels. It works correctly if I ask it to create a wav file with a single channel, so it seems to be something specific to the format/codec I am asking it to use. Is there a way to get it to do what I want, or is this a fundamental limitation?
[11:32:50 CET] <TarquinWJ> ffmpeg.exe -i myinput.wav -c:a libfdk_aac -profile:a aac_lc -ar 22050 -ac 1 -b:a 32k myoutput.m4a
[11:32:50 CET] <TarquinWJ> https://trac.ffmpeg.org/wiki/AudioChannelManipulation
[11:49:40 CET] Action: TarquinWJ meeting, sorry to have to ask and then disappear, I'll be back in about 20 minutes, if anyone has any helpful answers :)
[12:23:58 CET] <Mina> Hi, I am applying for FFmpeg as part of GSoC and I am almost done with the qualification task. I've some questions about the project proposal tho, is this the right place to ask them?
[12:26:59 CET] <durandal_1707> Mina: no, correct place is #ffmpeg-devel
[12:27:58 CET] <Mina> Thanks.
[12:31:10 CET] <TarquinWJ> sorry, back now, did anybody manage to answer my question?
[12:31:58 CET] <durandal_1707> no
[12:37:50 CET] <TarquinWJ> thanks
[13:41:08 CET] <hellyeah> hey anyone around?
[13:47:13 CET] <durandal_1707> me
[13:47:24 CET] <hellyeah> hmm good sup
[13:48:04 CET] <hellyeah> ffmpeg -v verbose -i rtsp://wowzaec2demo.streamlock.net/vod/mp4:BigBuckBunny_115k.mov myTestCam.m3u8 i have this command? How can i modify that code to send stream to website?
[13:58:50 CET] <DHE> iirc you can have the output be an HTTP URL and ffmpeg will attempt to execute PUT commands.. if that works with your hosting provider
[13:59:27 CET] <DHE> otherwise I recommend some form of remote filesystem mount... samba, sshfs, whatever's available
[14:11:00 CET] <Nacht> hellyeah: What exactly are you trying to do though ? Cause that's a VOD
[14:13:32 CET] <DHE> I'm assuming it's a test or demo...
[14:15:50 CET] <Nacht> Yeah but there's no point in streaming a VOD to a website, when you can just transcode/mux it and then upload it
[14:22:15 CET] <mort> Hey, how do I compile ffmpeg such that a program using libavcodec's avcodec_find_decoder(AV_CODEC_ID_H264) will chooses a hardware decoder ("h264_v4l2m2m") instead of the regular software decoder ("h264")?
[14:35:09 CET] <hellyeah> VOD?
[14:35:21 CET] <hellyeah> it is just example url
[14:35:29 CET] <hellyeah> i have ip cam
[14:35:42 CET] <hellyeah> i try got get rtsp stream and publis on the website
[14:38:02 CET] <DHE> hellyeah: VOD is a pre-rendered video. so you're just using bigbuckbunny as a test stream then?
[14:38:30 CET] <DHE> by "pre-rendered" I mean as in "not live/realtime"
[14:41:09 CET] <hellyeah> any live url available?
[14:44:07 CET] <hellyeah> hmm anyway
[14:47:06 CET] <mort> is there any way, using configure flags or environment variables or anything, to select the default video decoder?
[14:47:38 CET] <mort> I'm not invoking ffmpeg from the command line, but using a program which just uses avcodec_findi_decoder and uses whatever ffmpeg gives it
[14:48:49 CET] <mort> I would preferably like something a bit more robust than compiling with all codecs other than h264_v4l2m2m disabled
[14:50:01 CET] <mort> and there dosen't even seem to be a compile time option to disable the default software h264 decoder (at least without also disabling h264 decoding in general), unless I'm mistaken
[14:50:14 CET] <DHE> configure --disable-decoders=h264
[14:50:40 CET] <mort> and that disables only the decoder named "h264", and not all h264 decoders?
[14:50:42 CET] <DHE> but do watch the output of configure afterwards. there are dependencies between features. it's possible disabling the software decoder kills some hardware decoding features...
[14:50:57 CET] <DHE> configure --list-decoders
[14:51:00 CET] <DHE> to see what's available
[14:51:23 CET] <mort> it would be really nice to be able to just reorder them, to make h264_v4l2m2m the default
[14:51:38 CET] <mort> but from what I've seen in allcodecs.c, the only way to accomplish that would be to charnge the source code
[14:51:41 CET] <mort> change*
[14:52:21 CET] <DHE> you can enable and disable codecs from the configure script
[14:52:30 CET] <mort> yes, but not reorder
[14:56:40 CET] <hellyeah> ffmpeg -v verbose -i rtsp://184.72.239.149/vod/mp4:BigBuckBunny_175k.mov myTestCam.mp4
[14:56:47 CET] <hellyeah> what flag should i use here?
[14:59:10 CET] <hellyeah> that should work
[15:15:20 CET] <th3_v0ice> Does anyone have some sample code on how to properly generate HLS playlist using FFmpeg API? Setting the header options doesnt do anything. Thanks!
[15:30:18 CET] <barhom> 1080i source, Running yadif, then splitting the yadif (to be used in 2 different outputs) -filter_complex '[0:v]yadif,split=2[yadif1][yadif2]' \
[15:30:28 CET] <barhom> How do I also scale one to 720p and the other to something like 720x576 ?
[15:31:10 CET] <furq> TarquinWJ: that command works fine here (except it's "aac_low", not "aac_lc")
[15:31:44 CET] <kepstin> add ";[yadif1]scale=-2:720[yadif1_720]" for the 720, repeat slightly differently for the other stream
[15:31:55 CET] <kepstin> barhom: then map the scale output rather than the yadif output
[15:32:23 CET] <barhom> what does "-2" mean  here?
[15:32:44 CET] <furq> -1 is automatic to match aspect ratio, -2 is the closest mod2 value
[15:33:12 CET] <furq> also you probably want setdar/setsar in there for the 576p output
[15:36:59 CET] <kepstin> scale by default will preserve dar, which means you might sometimes get funky dar values if some rounding was performed.
[15:37:08 CET] <kepstin> you sometiems get funky sar values*
[15:39:19 CET] <furq> oh yeah nvm then
[15:39:50 CET] <furq> i just tested it and got 64:45, which is correct if you want to do pal things with this
[15:39:54 CET] <barhom> this worked;  -filter_complex '[0:v]yadif,split=2[yadif1][yadif2];[yadif1]scale=-2:720[yadif1_720];[yadif2]scale=720:576[yadif2_576]' \
[15:39:58 CET] <furq> if not then it doesn't matter anyway
[15:40:04 CET] <barhom> although Im not sure I should scale a 1080i source to 720x576
[15:40:35 CET] <furq> if it's for broadcast then sure
[15:40:51 CET] <barhom> trying to do a 3.5mbit "hd" and 1mbit "sd" out of the 1080i source
[15:41:06 CET] <furq> if it's for streaming then just scale to -2:576
[15:41:12 CET] <furq> or more likely -2:480 for 1mbit
[15:41:41 CET] <hellyeah> ffmpeg comment avoid playing video
[15:41:56 CET] <hellyeah> i have to stop the sreaming for watching video why is that=?
[15:42:23 CET] <kepstin> hellyeah: you're using an mp4 file - mp4 is not a streamable format, it can't be played until it's finished
[15:42:33 CET] <kepstin> use a different container if you want to play while recording.
[15:42:39 CET] <hellyeah> oh
[15:42:50 CET] <hellyeah> like ogg
[15:42:52 CET] <hellyeah> ?
[15:43:05 CET] <hellyeah> so that will work in live streaming
[15:43:06 CET] <hellyeah> ?
[15:43:23 CET] <kepstin> ogg doesn't have very wide codec support, so probably not great. maybe use mkv
[15:44:19 CET] <hellyeah> mkv hmm
[15:44:53 CET] <hellyeah> ffmpeg -i  rtsp://184.72.239.149/vod/mp4:BigBuckBunny_175k.mov -f mkv -vcodec libx264 -preset fast -profile:v main -acodec aac example.mp4 -hide_banner smt like that?
[15:45:13 CET] <kepstin> hellyeah: just change the file extension to .mkv
[15:45:26 CET] <hellyeah> ok
[15:45:29 CET] <kepstin> don't need -f, it's autodetected from extension
[15:46:00 CET] <TarquinWJ> furq: thanks for that reply! Ì'll see if that produces the output we're looking for
[15:46:04 CET] <hellyeah> ffmpeg -i  rtsp://184.72.239.149/vod/mp4:BigBuckBunny_175k.mov  -vcodec libx264 -preset fast -profile:v main -acodec aac example.mkv -hide_banner
[15:46:10 CET] <hellyeah> let me tru that
[15:50:10 CET] <hellyeah> mkv is not playyable on web
[15:52:26 CET] <pmjdebru1jn> hellyeah: aren't you looking for dash then?
[15:54:06 CET] <hellyeah> actually i try to play vid on web
[15:54:22 CET] <pmjdebru1jn> play or stream?
[15:54:38 CET] <hellyeah> play
[15:54:42 CET] <pmjdebru1jn> and "the web" means... which device/browser
[15:55:10 CET] <pmjdebru1jn> rtsp sortof seems to imply stream?
[15:55:59 CET] <hellyeah> got rtsp on command line and convert it to mkv
[15:56:06 CET] <hellyeah> then try to play mkv on web
[15:56:16 CET] <hellyeah> <video>
[15:56:16 CET] <hellyeah>     <source src="example.mkv" type="video/mp4">
[15:56:16 CET] <hellyeah> </video>
[15:56:19 CET] <hellyeah> with that
[15:56:35 CET] <pmjdebru1jn> that doesn't make sense in any case
[15:56:42 CET] <pmjdebru1jn> video/mp4 implies an .mp4 file
[15:57:10 CET] <pmjdebru1jn> https://github.com/google/shaka-player
[15:57:13 CET] <hellyeah> from stackoverflow https://stackoverflow.com/questions/21192713/how-to-playback-mkv-video-in-web-browser
[15:57:53 CET] <pmjdebru1jn> it's still wrong
[15:58:00 CET] <pmjdebru1jn> obviously so :)
[15:58:30 CET] <pmjdebru1jn> for streaming H264 content you need to look into HLS/DASH
[15:58:39 CET] <pmjdebru1jn> or transcode to WebM/VP8/VP9
[15:58:52 CET] <pmjdebru1jn> but that excludes Apple devices
[15:59:25 CET] <pmjdebru1jn> using an .mpd extension might suffice
[15:59:32 CET] <pmjdebru1jn> https://developer.mozilla.org/en-US/docs/Web/HTML/DASH_Adaptive_Streaming_for_HTML_5_Video
[15:59:50 CET] <pmjdebru1jn> I meant in the ffmpeg cmdline
[16:00:08 CET] <pmjdebru1jn> ffmpeg might recognize you want a dash compatible file
[16:01:30 CET] <hellyeah> what command could be can you help me with that?
[16:01:46 CET] <hellyeah> oh
[16:01:50 CET] <hellyeah> let me check
[16:04:26 CET] <pmjdebru1jn> hellyeah: ffmpeg -i whatever -c:v libx264 -profile:v main -preset:v fast -c:a aac output.mpd
[16:04:54 CET] <pmjdebru1jn> might be worth trying
[16:05:00 CET] <hellyeah> mpd is playyable on web?
[16:05:12 CET] <pmjdebru1jn> it might be
[16:05:18 CET] <pmjdebru1jn> mpd should be MP4+Dash
[16:06:09 CET] <kepstin> note that the browser can't play it with just a video tag, you'll need a javascript dash player library
[16:07:00 CET] <pmjdebru1jn> like shaka-player :)
[16:07:45 CET] <hellyeah> or hls?
[16:08:00 CET] <kepstin> some mobile browsers can play hls natively
[16:08:08 CET] <kepstin> and maybe desktop safari as well, not sure
[16:09:31 CET] <hellyeah> i need to go now
[16:09:35 CET] <hellyeah> thanks for helping
[16:09:43 CET] <hellyeah> i will be hanging there
[16:09:46 CET] <hellyeah> good day
[19:01:56 CET] <transcodeine> trying to figure out how to get  yuv422p10le to yuv420p using ffmpeg but can't seem to figure out how to do it- have tried everything to no avail.  i presume it's because i need the 10 bit static to x264 and the 8 bit static to 264 which can't be referenced concurrently
[19:06:42 CET] <The-Present> Hi FFMPGers. Need some help and may be an advice.
[19:06:51 CET] <The-Present> What is connection reset by peer?
[19:07:12 CET] <The-Present> if i stream 20 streams , couple of them fails with connection reset by peer. The rest goes through.
[19:08:06 CET] <The-Present> JEEB, Hi. Can you help in this?
[19:09:37 CET] <durandal_1707> do not leech
[19:09:59 CET] <The-Present> durandal_1707, ?  sorry didn't get you?
[19:19:15 CET] <DHE> transcodeine: the latest git version of ffmpeg+x264 supports all bit depths in a single binary....
[19:19:30 CET] <DHE> note that it's only the output to h264 that is bit-restricted in older versions
[19:22:16 CET] <kepstin> transcodeine: also note that ffmpeg can always decode all bit depths. The x264 version only affects what it can encode.
[19:29:32 CET] <iive> The-Present, connection reset by peer, usually means that the client on the other side has closed the connection
[19:32:26 CET] <The-Present> iive: Thank you very much for this answer. My next query was "if we have 20 streams and couple of them fails with this error, what is the most common cause". Remote RTMP server is same for all 20 streams.
[19:33:25 CET] <iive> The-Present, no idea, sorry.
[19:33:59 CET] <The-Present> iive, ok thank you. :)
[19:34:49 CET] <DHE> best guess is that ffmpeg will open them all one-by-one and maybe some are timing out?
[19:40:51 CET] <transcodeine> DHE- AMAZING
[19:40:56 CET] <transcodeine> i had no idea.
[19:41:03 CET] <transcodeine> did i just get lucky? that must have been recent?
[19:44:18 CET] <DHE> transcodeine: late december 2017 iirc
[19:44:44 CET] <furq> transcodeine: that shouldn't affect decoding though
[19:44:52 CET] <furq> the builtin decoder will decode 8 and 10-bit
[19:45:03 CET] <furq> it's just libx264 that (traditionally) has a specific bit depth
[19:45:34 CET] <The-Present> iive, is there a way to set max speed 1x for ffmpeg streaming without encoding. We are using -c copy to copy codecs and no encoding is involved.
[19:46:07 CET] <sfan5> use -re
[19:46:09 CET] <iive> i'm not really familiar with streaming at all.
[19:47:30 CET] <The-Present> sfan5, we are using -re.
[19:47:41 CET] <The-Present> do we have to pass in some value for -re
[19:47:46 CET] <sfan5> nope
[19:47:48 CET] <The-Present> iive, thanks
[19:47:54 CET] <sfan5> is it still faster than 1x with -re?
[19:48:13 CET] <furq> The-Present: if you have multiple inputs, you need -re on all of them
[19:49:39 CET] <The-Present> ok we use fifo and it was 22.5x
[19:50:01 CET] <The-Present> i am not very good at ffmpeg but i love it and testing it like crazy.
[19:50:24 CET] <The-Present> We always encounter 3 errors always when streaming in bulk using -c copy option
[19:50:25 CET] <transcodeine> DHE- thanks man- much appreciated. i'll pull that now.
[19:50:37 CET] <The-Present> furq, no we have one input
[19:52:07 CET] <transcodeine> DHE- i don't see that in the Changelog but will try anyway
[19:52:32 CET] <The-Present> we have set connection options like "-reconnect 1 -reconnect_at_eof 1 -reconnect_streamed 1 -reconnect_delay_max " to not get reset errors
[19:52:36 CET] <The-Present> but still we get them
[19:53:33 CET] <transcodeine> furq- encoding
[19:54:20 CET] <The-Present> furq, do you have happen to know if we use 1 input and multiple outputs for rtmp without encoding. Copying audio and video codec instead of re-encoding. can tee muxer be used without encoding?
[19:54:30 CET] <transcodeine> 10 > 8 bit
[19:54:38 CET] <transcodeine> same binary.
[19:56:29 CET] <JEEB> The-Present: yes
[19:57:35 CET] <The-Present> JEEB, Hi ...
[19:59:54 CET] <The-Present> Do you know if its possible to have single input -> multiple output with tee muxer and no encoding?
[20:01:08 CET] <DHE> probably. why not?
[20:01:19 CET] <transcodeine> - 10-bit depth support in native utvideo decoder
[20:01:22 CET] <transcodeine> guess that's it?
[20:01:54 CET] <furq> transcodeine: are you trying to convert from 10-bit to 8-bit
[20:02:00 CET] <furq> or did i misread your question
[20:02:11 CET] <transcodeine> furq yes sir
[20:02:20 CET] <furq> well yeah you should just need an 8-bit x264
[20:02:48 CET] <furq> the builtin decoder will handle both bit depths
[20:10:22 CET] <DHE> but you'll need a -pix_fmt parameter to convert to the 8bit version of the colour depth. x264 will pick up on that and produce 8bit output (assuming the mix depth builds)
[20:10:41 CET] <ggggggg> hello
[20:11:16 CET] <transcodeine> DHE- yeah i tried that but it just ignored it
[20:11:24 CET] <transcodeine> -pix_fmt yuv420p
[20:11:25 CET] <furq> you shouldn't even need to set -pix_fmt
[20:11:32 CET] <furq> pastebin your full command and output
[20:11:50 CET] <transcodeine> ok
[20:13:01 CET] <furq> [format @ 0x80e036100] auto-inserting filter 'auto_scaler_0' between the filter 'Parsed_null_0' and the filter 'format'
[20:13:04 CET] <furq> [auto_scaler_0 @ 0x80e0361a0] picking yuv420p out of 8 ref:yuv420p10le alpha:0
[20:13:19 CET] <furq> so yeah you probably have some other issue
[20:13:34 CET] <furq> maybe a build with 10-bit x264
[20:13:47 CET] <transcodeine> yes exactly
[20:13:54 CET] <transcodeine> yuv420p10le > yuv420p
[20:15:03 CET] <transcodeine> https://pastebin.com/6Vjf1AHf
[20:15:20 CET] <furq> i sort of wanted the output as well
[20:15:36 CET] <transcodeine> ok
[20:15:36 CET] <furq> but if it's just ignoring you and encoding yuv420p10le then you need an ffmpeg with 8-bit x264
[20:15:45 CET] <transcodeine> definitely ignoring
[20:16:22 CET] <furq> also you're setting -level twice and you're setting the video bitrate to 218 bits per second
[20:17:02 CET] <transcodeine> that could just be an artifact or something- yeah the command line is messy at the moment
[20:17:24 CET] <transcodeine> i'm just trying to get it to take the x264 opts
[20:19:11 CET] <transcodeine> i'm compiling from scratch right now.
[20:19:15 CET] <sfan5> I don't think multiple -x264-params work
[20:19:23 CET] <transcodeine> it was a static bin i got- clearly an issue
[20:20:01 CET] <sfan5> also you're setting some encoding options after your output filename
[20:42:50 CET] <transcodeine>    Stream #0:1(eng): Video: h264 (libx264) (avc1 / 0x31637661), yuv420p, 1280x720 [SAR 1:1 DAR 16:9], q=-1--1, 0.02 fps, 11988 tbn, 59.94 tbc (default)
[20:42:52 CET] <transcodeine> ^^ bam
[20:43:03 CET] <transcodeine> old binary :\
[20:43:05 CET] <transcodeine> thanks folks
[20:45:17 CET] <transcodeine> it's also  encoding two fold faster
[20:51:02 CET] <lyncher> I'm having this error when doing trace_headers to a h264 file: rbsp_stop_one_bit out of range
[20:51:15 CET] <lyncher> what does this means? any hint?
[20:51:59 CET] <JEEB> rbsp_stop_one_bit is probably a field in H.264
[20:52:00 CET] <jkqxz> That syntax element had a value it isn't allowed to.  Probably your file is invalid.
[20:52:06 CET] <JEEB> so yes
[20:52:08 CET] <JEEB> what jkqxz says
[20:52:28 CET] <JEEB> of course it's possible that the code is bugged so if you want someone else to look at your file you can post it somewhere sane
[20:52:35 CET] <JEEB> and make a ticket against it on the trac
[20:53:21 CET] <lyncher> I'm betting on a bug.... that file is being produced by an broadcast grade HW encoder
[20:53:33 CET] <jkqxz> Post the file?
[20:53:59 CET] <lyncher> what service do you usually use to share a file (30M)
[20:54:03 CET] <lyncher> ?
[20:54:16 CET] <JEEB> there's some sort of uploader on ffmpeg.org that I have no idea about credentials for
[20:54:18 CET] <jkqxz> <https://0x0.st/>
[20:54:21 CET] <JEEB> or that yes
[20:54:26 CET] <JEEB> you can just POST to it
[20:54:30 CET] <JEEB> with curl or something
[20:54:32 CET] <foobar_user68195> Hello. I have a problem with putting audio from one video to another one as a default audio track (but first I'm doing some manually-adjusted-for-fps-differences trimming). https://pastebin.com/hNPyEd62 - commands; https://pastebin.com/QxNZjP3c - unintelligible error message
[20:55:16 CET] <foobar_user68195> (I'm trying to do it with one command because I'm HDD constrained while RAM and CPU are bored at night.)
[20:55:48 CET] <JEEB> lyncher: for the reference I have seen very expensive encoders spit out slightly incorrect bit streams, but I don't think anyone can say anything until the sample is seen :)
[20:57:14 CET] <lyncher> I've uploaded a sample to: http://0x0.st/sBMs.ts
[20:57:46 CET] <JEEB> jkqxz: can you take a look at it specifically, or shall I recommend he make a bug on trac?
[20:57:59 CET] <jkqxz> I'll have quick look now.
[20:58:03 CET] <JEEB> cheers
[20:58:08 CET] <lyncher> thanks!
[21:02:51 CET] <jkqxz> It's invalid.  The AUDs are missing the RBSP trailing bits - "00 01 09 e0 00".
[21:03:32 CET] <jkqxz> They've kindof failed at AUDs, too.  They all have primary_pic_type == 7 (i.e. "this AU can contain all slice types"), which is kindof useless.
[21:05:57 CET] <lyncher> thanks for your quick reply. I'll check what generated that stream
[21:08:25 CET] <foobar_user68195> Soooo... can anyone tell what's wrong with that filter_complex?
[21:08:57 CET] <foobar_user68195> It's been three days at documentation, google, stackexchange and trying out and this is my best result (-.-)'
[21:13:13 CET] <lyncher> dveo encoder
[21:13:54 CET] <lyncher> is there any way to fix those streams (maybe a bsf)?
[21:17:36 CET] <jkqxz> Apply <https://0x0.st/sBMK.diff> and then run it through -bsf:v 'h264_metadata=aud=remove,h264_metadata=aud=insert'.
[21:18:07 CET] <jkqxz> (I.e. allow the errorneous value, remove the invalid AUDs, write it again with correct AUDs.)
[21:22:01 CET] <Guest23959> anyone know how to show the channel assignments for a 5.1 wav file as it is embeded in the wav EXT info?
[21:24:01 CET] <lyncher> it worked.....!
[21:24:13 CET] <lyncher> just applied the patch a recompiled
[21:24:34 CET] <lyncher> I've didn't add the bsf metadata filters
[21:25:20 CET] <lyncher> I suppose that those filters will fix the AUD units. the pach allow ffmpeg to correctly deal with those files. right?
[21:28:05 CET] <jkqxz> Well, "correctly".  They are wrong, but it's pretty obvious what the encoder intended.
[21:28:57 CET] <jkqxz> Without the h264_metadata instances it will be able to run trace_headers on it (and you will be able to see the incorrect rbsp_stop_one_bit value in the output), but the stream isn't change.
[21:29:33 CET] <transcodeine> is there any added advantage at all in ffmpeg to running more than 16 threads?  i have 72 vcpus available and ideally would like to run one thread per.
[21:30:07 CET] <transcodeine> i'm getting 4x doing some fairly heavy transcoding so i'm not complaining but wondering if i could speed it up even more..
[21:35:25 CET] <th3_v0ice>  Does anyone have some sample code on how to properly generate HLS playlist using FFmpeg API? Setting the header options doesnt do anything. Thanks!
[21:45:57 CET] <lemonchicken> I get mjpeg stream errors, "Found EOI before any SOF, ignoring", and "No JPEG data found in image", when decoding mjpeg stream off an avermedia game capture card, can anyone point out to where i should to fix this?
[22:00:47 CET] <alexpigment> there's an avermedia capture card that capture MJPEG?
[22:00:48 CET] <alexpigment> weird
[22:01:49 CET] <alexpigment> anyway, copy the console output from your ffmpeg command line to the end of the output
[22:01:56 CET] <alexpigment> then put it in a pastebin
[22:02:05 CET] <alexpigment> someone here will possibly know what's going on
[22:13:11 CET] <lyncher> jkqxz: will this patch going to git-head or is just a custom fix for this file?
[22:20:59 CET] <jkqxz> It's purely a workaround for the error in that file.
[22:38:40 CET] <caimlas> hello, I'm trying to figure out how to compile ffmpeg with full nvidia/cuda support. While using --enable-libnpp I'm told "ERROR: libnpp not found" after having already installed the nvidia cuda packages provided by nvidia (on ubuntu 16.04) and appear to have numerous /usr/local/cuda/lib64/libnpp* files. I've followe the (brief) instructions here: https://developer.nvidia.com/ffmpeg
[22:39:45 CET] <caimlas> here his how I'm trying to configure: https://pastebin.com/ksxYdayg
[22:40:11 CET] <caimlas> is anyone able to point me in the right direction to get --enable-libnpp to work?
[22:55:04 CET] <jkqxz> caimlas:  Look in ffbuild/config.log to find what actually went wrong.  Maybe it isn't looking in the right place or is looking for something you don't have.
[22:58:08 CET] <caimlas> jkqxz, interestingly, one is not generated.
[22:58:48 CET] <caimlas> urg nevermind, that
[22:59:50 CET] <caimlas>  /usr/bin/ld: cannot find -lnppi
[23:02:02 CET] <caimlas> jkqxz, I can paste the full log somewhere; reading through configure now to try to figure out what that means, though
[23:08:08 CET] <jkqxz> Well, that means libnppi wasn't found anywhere in the paths you provided to it.  Probably you are either missing it or you want to provide another -L option for where it is.
[23:27:11 CET] <barhom> https://pastebin.com/TdXaFrRe < I feel this is the wrong way to set different bitrates for the two outputs (720p, 576). Can somebody educate me here?
[23:27:41 CET] <barhom> The command runs but it seems both outputs are transcoded to 1000k
[23:35:44 CET] <furq> barhom: [output1 options] output1 [output2 options] output2
[23:36:06 CET] <furq> -b:v:1 in your command is selecting the second stream in the first output, not the second output
[23:37:40 CET] <furq> actually nvm that looks correct for hls variants
[23:38:14 CET] <barhom> master playlists is quite new in ffmpeg, I read this from some doc
[23:38:17 CET] <furq> yeah it is
[23:38:24 CET] <furq> maxrate and bufsize probably need stream specifiers
[23:38:34 CET] <barhom> how would I do that?
[23:38:42 CET] <furq> presumably -maxrate:v:1
[23:38:53 CET] <barhom> Ill try that
[23:39:06 CET] <barhom> but is the thoght process correct here with filters?
[23:39:15 CET] <barhom> apply yadif to source, split it, scale it, encode it
[23:39:23 CET] <furq> looks fine
[23:39:39 CET] <barhom> or would I apply yadif after scaling (2 yadifs)
[23:39:46 CET] <furq> no you always deinterlace before scaling
[23:40:28 CET] <furq> my only suggestion would be that 1024x576 is pretty big for an sd variant, especially at 1mbit
[23:40:32 CET] <furq> i'd probably go with -2:480
[23:41:05 CET] <barhom> I'll tweak and see what looks good, thanks
[23:41:21 CET] <barhom> hmm, what about encoding the audio once
[23:41:27 CET] <barhom> wonder how much cpu time I would save
[23:41:36 CET] <furq> you're already doing that
[23:41:49 CET] <furq> actually nvm no you're not
[23:43:29 CET] <furq> i feel like there's a way to do this that doesn't involve the tee muxer
[23:44:13 CET] <barhom> where am I using tee? (split?)
[23:44:17 CET] <furq> you're not
[23:44:34 CET] <furq> tee would let you encode once
[23:44:49 CET] <furq> but it also massively overcomplicates things and i have no idea if it works with hls
[23:45:46 CET] <barhom> cant I use a filter to split the audio much like I do with the yadif output?
[23:46:09 CET] <furq> you can but that won't do anything useful
[23:46:18 CET] <furq> the filters are applied before encoding
[23:46:52 CET] <furq> the aac encoding is a drop in the ocean compared to two x264 encodes anyway
[23:47:17 CET] <furq> i wouldn't worry about it unless you really need an extra 2%
[23:47:22 CET] <furq> if that
[23:47:58 CET] <barhom> https://pastebin.com/uwkpqtR0 < this works now
[23:48:08 CET] <barhom> minus the double aac encodes
[23:48:22 CET] <barhom> thanks for the -maxrate:v:X
[23:49:17 CET] <furq> the easiest solution is to just make -b:a:0 96k so you don't feel as bad about it
[23:49:51 CET] <furq> that's only half joking, 64k is probably a bit low for 720p
[23:50:31 CET] <barhom> right
[23:51:03 CET] <barhom> tomorrows task will be how to do both HLS and DASH in one go
[23:51:37 CET] <furq> idk if there's much point really
[23:51:42 CET] <furq> even youtube only uses hls for livestreaming
[23:52:03 CET] <barhom> did not know that
[23:52:23 CET] <barhom> what about doing the new fancy fmp4 instead of ts
[23:53:23 CET] <furq> idk if ffmpeg supports that with the hls muxer yet
[23:53:30 CET] <furq> but it's not really worth it anyway
[23:53:49 CET] <furq> i assume ts is better supported on stuff that has native hls support
[23:54:25 CET] <furq> the problem with dash for livestreaming is that apple refuse to support it
[23:54:32 CET] <furq> so youtube is just hls for live and then dash or http for vod
[23:57:45 CET] <barhom> Any idea if I can keep the 1080i source as-is? Would probably break apples rules though
[23:57:55 CET] <barhom> they dont want interlaced segments
[23:58:14 CET] <barhom> source (1080i), 720p, 480p
[00:00:00 CET] --- Fri Mar 23 2018


More information about the Ffmpeg-devel-irc mailing list