[Ffmpeg-devel-irc] ffmpeg.log.20180102

[burek]

[01:20:34 CET] <Johnjay> not that this matters
[01:20:51 CET] <Johnjay> but i found youtube-dl seems better at downloading mp2t streams and encoding them than ffmpeg alone does
[01:21:25 CET] <Johnjay> in this case I gave the .m3u8 file as the input and ffmpeg stopped after a minute or so, but youtube-dl seems to have finished
[01:55:29 CET] <Johnjay> yeah spoke too soon, nevermind
[01:55:39 CET] <Johnjay> youtube-dl just went longer is all
[03:00:31 CET] <voldial> Can I specify the mkv Segment UID? ffmpeg -y -i 'file:test.mp4' -i 'file:test.webm' -c copy -map 0:v:0 -map 1:a:0 'file:test.mkv'
[03:07:23 CET] <voldial> hm. MODE_WEBM https://github.com/FFmpeg/FFmpeg/blob/master/libavformat/matroskaenc.c#L1923
[03:11:48 CET] <DHE> $ ffmpeg -h muxer=matroska   # shows what options the MKV muxer accepts. yeah, nothing there.
[03:31:13 CET] <voldial> ah. thanks
[08:01:54 CET] <jpsollie> hello everyone, can anybody help me finding the part of ffmpeg where the h264_vaapi profile -> features mapping is done?
[08:03:12 CET] <jpsollie> i'm looking for this part because my AMD Fiji and Tonga cards both output a stream which uses separate color planes, but it is unsupported
[08:03:38 CET] <jpsollie> so I want to force it to disable this part
[08:04:00 CET] <jpsollie> sorry, feature
[09:03:13 CET] <ESphynx> please help guys... how do I just select another audio stream?
[09:10:10 CET] <illegal> for what? anyway I think the option is -map #
[09:18:09 CET] <ESphynx> right I was just struggling to understand that I nede to -map 0:0 for the video, then -map 0:4 for my audio, both options together...
[09:18:23 CET] <ESphynx> though the man page explained it :P
[10:15:49 CET] <kingsley> Could melt crash with a segfault if it passes acodec=vorbis instead of acodec=libvorbis to ffmpeg?
[10:46:42 CET] <Spring> can ffmpeg copy streams from m2ts files for trimming?
[10:47:14 CET] <Spring> basically wondering what would be best for creating short clips of rips for encoding test purposes
[10:49:19 CET] <sfan5> it can yes
[10:49:31 CET] <pmjdebruijn> Spring look int o-ss -t
[10:49:54 CET] <pmjdebruijn> and -c:v copy -c:a copy
[10:50:17 CET] <pmjdebruijn> although I do wonder whether you can use both
[10:50:24 CET] <pmjdebruijn> both regard to iframes and such
[10:50:27 CET] <pmjdebruijn> with*
[11:32:02 CET] <kiranbsravi> Hi. I am trying to mux audio and video to a webm file using libavformat. For video stream, when i call av_interleaved_write_frame() with AVPacket,  write_packet() -> mkv_write_flush_packet() in matroskaenc.c is called.
[11:33:36 CET] <kiranbsravi> there MatroskaMuxContext.cur_audio_pkt is unrefed, and it is throwing "malloc: *** error for object 0x7fda5f816200: pointer being freed was not allocated"  error. implying the AVBuffer's data was never allocated inside AVPacket.
[11:34:37 CET] <kiranbsravi> i haven't  freed any modules. Is there something i am missing ?
[11:49:35 CET] <BtbN> Sounds like you never zero-initialized the struct, or did not allocate it properly
[11:57:02 CET] <kiranbsravi> Which struct u mean ?
[12:05:07 CET] <strfyr> i'm wondering, can i do something to improve the avcodec h264 decoder against packet loss?
[12:27:59 CET] <DHE> while h264 does support a form of forward error correction for that sort of thing, I don't recall it being supported by ffmpeg
[13:09:03 CET] <pmjdebruijn> DHE: the other issue is whether your client supports it
[13:09:18 CET] <pmjdebruijn> DHE: also look into whether libx264 supports it
[13:09:37 CET] <pmjdebruijn> https://www.ietf.org/rfc/rfc5109.txt it seems to be part of RTP as opposed to h264
[17:22:45 CET] <dongs> am i not supposed to be allowed to use -vf/crop filter when doing cuvid->nvenc encode?
[17:22:58 CET] <dongs> does the decoded stuff never leave the GPU or something
[17:26:16 CET] <dongs> ok looks like it
[17:26:26 CET] <dongs> ill jus use dxva instead
[17:28:58 CET] <Chloe> I'm trying to 'tile https://0x0.st/sPKc.png
[17:29:00 CET] <Chloe> fk
[17:29:15 CET] <dongs> gg
[17:29:19 CET] <Chloe> hate this keyboard, enter key trash
[17:29:28 CET] <dongs> you must be on a japanese keyboard
[17:29:30 CET] <dongs> or a mac
[17:31:26 CET] <Chloe> I'm trying to 'tile' an image (https://0x0.st/sPKc.png), there seems to be two ways to do it. The first using the tile filter and looping the image with `ffmpeg.exe -loop 1 -t 1 -i in.png -filter_complex tile=50x50 out.png`, but it seems to output a broken image https://0x0.st/sPKA.png
[17:31:58 CET] <Chloe> the second is using hstack + vstack, though im not sure how to use it exactly (still need to read the docs for it)
[17:33:08 CET] <durandal_1707> Chloe: thats just black for missing frames
[17:33:39 CET] <durandal_1707> for single image you need to loop 50x50 frames
[17:33:41 CET] <Chloe> I actually have no idea how to use ffmpeg, nice. Guess I need to increase loop time
[18:32:36 CET] <Captain_Haddock> I'm looking to add a fade in/out options to text via drawtext. Examples of the net involve doing this by splitting the input file, adding the text to one, overlaying it, and manipulating its alpha channel. These examples are however from a few years ago. Is this still the case with the latest versions?
[19:09:16 CET] <pgorley> hi, what's the difference between d3d11va and d3d11va2 in the configure script?
[19:12:23 CET] <pgorley> ./configure --list-hwaccels shows both h264_d3d11va and h264_d3d11va2
[20:04:42 CET] <d4re> hey im having an issue with concat. the audio drops after a while and the logs are populated with [hls @ 0x25fe120] Delay between the first packet and last packet in the muxing queue is 10010000 > 10000000: forcing output
[20:05:58 CET] <d4re>  command: -loglevel debug -hide_banner -re -f concat -safe 0 -fflags +genpts -async 1 -stream_loop -1 -i /trailerparkboys.txt -y -framerate 24 -c:a libfdk_aac -b:a 128k -c:v copy -pix_fmt yuv420p -level 3.2 -maxrate 2M -bufsize 6M -crf 18 -r 24 -g 72 -f hls -segment_list_flags +live -hls_flags +discont_start -hls_time 10 -hls_wrap 20 -s 1280x720 /srv/stream.m3u8
[20:23:41 CET] <}8]> hey fellas. does anyone know if support for the mkv crop flag is on the horizon?
[20:24:10 CET] <durandal_1707> whats mkv crop?
[20:25:11 CET] <sfan5> probably PixelCrop{Top,Bottom,Left,Right} (https://www.matroska.org/technical/specs/index.html)
[20:26:54 CET] <furq> }8]: if this is h264 then you can set the h264 crop values now, which should be better supported by players
[20:27:04 CET] <durandal_1707> isnt that exported as side data?
[20:27:21 CET] <furq> !bsf h264_metadata
[20:27:21 CET] <nfobot> furq: http://ffmpeg.org/ffmpeg-bitstream-filters.html#h264_005fmetadata
[20:29:29 CET] <}8]> furq - mkv crop settings dont require re-encoding the video. its very convenient & user-friendly to just set the crop values in the container
[20:29:43 CET] <furq> you don't need to reencode to apply a bsf
[20:30:08 CET] <furq> i'm not saying ffmpeg shouldn't support it but if it is h264 then it's better to change it in the sps
[20:30:26 CET] <furq> all h264 decoders should support that whereas mkv crop values aren't always obeyed by players
[20:31:07 CET] <}8]> theres always a player somewhere that doesnt work right but ive found most do
[20:31:38 CET] <}8]> crippled players shouldnt set the bar imo
[20:33:04 CET] <}8]> theres still tons of mpeg2 out there, and a lot of people remuxing from avi/mp4 into mkv now that mkv is widely supported and superior to those containers
[20:35:44 CET] <alexp> }8] is this just so that kodi will show 2.39:1 rather than 16:9?
[20:36:53 CET] <alexp> rather, will show it in the metadata properties?
[20:41:47 CET] <}8]> i dont know about that but so any player using ffmpeg has proper video when mkv crop is used
[20:42:34 CET] <sfan5> i'm fairly sure cropping would be responsibility of the API user, so support in just ffmpeg would mean nothing
[20:42:50 CET] <alexpigment> }8] i figured i'd ask. I saw some thread about people wanting kodi to say a video was [whatever] aspect ratio in their library rather than 16:9. it seemed like a very superficial and weird request
[20:43:20 CET] <}8]> easy way to get rid of unwanted borders, which in some cases arent even equal on all sides
[20:43:24 CET] <alexpigment> anyway, unless i had a display wider than 16:9, the crop feature wouldn't really do much for me
[20:44:27 CET] <alexpigment> }8] I've definitely seen that. I guess I'd be more concerned with non-integer scaling when cropped. my completely irrelevant 2 cents on the matter ;)
[20:54:36 CET] <furq> }8]: oh right, you mean in the demuxer
[20:54:49 CET] <furq> that makes more sense then i guess
[21:00:48 CET] <}8]> yeah furq :)
[00:00:00 CET] --- Wed Jan  3 2018