[Ffmpeg-devel-irc] ffmpeg.log.20170210

Sat Feb 11 03:05:01 EET 2017

[00:00:14 CET] <roasted_> bingo. nice. is there a vaapi flag that you know of? I assume something like --hwdec=vaapi or -hwaccel-vaapi ?
[00:00:34 CET] <DHE> dunno. chances are pretty good any ffmpeg input and filtering paramters will work in ffprobe
[00:00:36 CET] <DHE> err, ffplay
[00:02:34 CET] <Mista-D> Any way to edit scan type in ts or mp4 file please, without transcoding?
[00:04:09 CET] <kepstin> Mista-D: like, switch a stream between marked as interlaced vs. progressive?
[00:04:20 CET] <kerio> how would that even work
[00:04:35 CET] <kerio> fields have half the vertical resolution
[00:05:18 CET] <kepstin> most codecs encode a pair of fields together as a frame, but yeah - there's usually special encoding modes used for interlaced stuff to make it less inefficient
[00:05:57 CET] <Mista-D> kepstin; yes. it seems to stop HLS playback on iOS devices. While video is perfectly watchable.
[00:06:12 CET] <Mista-D> kepstin:
[00:06:47 CET] <kepstin> Mista-D: if the video's using an interlaced encoding mode, you'll probably have to re-encode so the ios decoder doesn't fail
[00:07:44 CET] <DHE> you might be able to run the yadif deinterlacer which is supposed to be smart enough to do nothing to already-progressive content. not sure if it's smart enough to deal with changing content
[00:09:22 CET] <kepstin> if the video is already progressive, it should be as simple as re-encoding with the filter setfield=mode=prog
[00:10:31 CET] <kepstin> like, if the video is progressive frames stored with 2:2 pulldown in interlaced video
[01:29:15 CET] <spiderkeys> Is anyone familiar with the new 3.1+ decode APIs? I'm trying to decode and use the latest frame I receive from a udp stream, but it feels like I am still not depleting the available buffer and my video is getting behind. Also I think I am not correctly initting and freeing resources. My code, if anyone could take a quick glance: http://coliru.stacked-crooked.com/a/01d1c87812393fd5
[01:38:46 CET] <roasted_> anybody know offhand what version of ffmpeg vaapi support came?
[02:00:19 CET] <llogan> you could use "git log" to find out
[02:00:54 CET] <llogan> or look at Changelog
[04:57:09 CET] <xtina> hey guys, i have a stupid question. i've written some ffmpeg commands that work from command line but not when i put them in a shell script
[04:58:12 CET] <xtina> http://pastebin.com/fxmPjZT0
[04:58:20 CET] <xtina> that works if i run it from the command line
[04:58:28 CET] <xtina> but from the shell, it looks like ffmpeg is never opening the video pipe
[05:28:51 CET] <xtina> i keep getting this message in my ffmpeg log, what does it mean?
[05:28:51 CET] <xtina> M[flv @ 0x301e2c0] Delay between the first packet and last packet in the muxing queue is 10100000 > 10000000: forcing output
[05:43:29 CET] <xtina> furq: do you recognize that error? ^^
[05:45:03 CET] <xtina> also, i'm trying to understand what the reported fps in the ffmpeg output means. i'm getting a steady 10fps (which is my target) but sometimes my youtube reports that the stream it's receiving is low data
[05:45:12 CET] <xtina> even though the fps reported by ffmpeg never changes, it's always at 10
[05:46:24 CET] <xtina> i do notice the reported bitrate dropping once the 'Delay between the first packet and last packet in the muxing queue' errors start coming
[05:46:34 CET] <xtina> and then i get that error every second?
[05:52:22 CET] <xtina> llogan: do you happen to recognize that error?
[07:36:21 CET] <Guest55> Is there a know issue with using vaapi and setting a -level
[07:36:34 CET] <Guest55> I'm seeing it keep the old level
[07:36:59 CET] <Guest55> ffmpeg -hide_banner -hwaccel vaapi -hwaccel_output_format vaapi -vaapi_device /dev/dri/renderD128 -i a.mp4 -c copy -vf 'format=nv12|vaapi,hwupload' -c:v h264_vaapi -level 40 -y b.mp4
[11:33:09 CET] <abhishek_> hi
[11:33:34 CET] <Guest80987> while using CPU usage is very high
[11:33:39 CET] <Guest80987> what to tune ?
[11:39:53 CET] <thebombzen> if you're talking about tuning libx264 encoding, that doesn't affect cpu usage
[11:42:22 CET] <furq> unless the source is realtime
[11:49:59 CET] <acamargo> good morning everyone. I've been where this number comes from. https://github.com/FFmpeg/FFmpeg/blob/master/libavformat/udp.c#L135 why 7*4096?
[11:50:14 CET] <acamargo> ...I've been thinking where...
[12:09:28 CET] <roasted_> I was reading that ffplay doesn't support near the options ffmpeg does. I'm trying to test a stream from a surveillance camera (rtsp) but I'd rather see if I can use ffmpeg as the recording on the server leverages ffmpeg (just so I can tinker as close to what the server uses as possible). Is it not possible to ffmpeg rtsp:// a stream without recording? I'd like to see it live but capture nothing in the process.
[12:10:46 CET] <furq> pipe it to ffplay or mpv
[12:11:50 CET] <roasted_> furq: the second I hit the enter key that thought popped into my head. ha.
[12:11:54 CET] <roasted_> looking into that now. :)
[12:12:07 CET] <furq> -i rtsp://foo -c copy -f mpegts - | mpv -
[12:14:19 CET] <roasted_> furq: where do I overnight the thanksbeer? :P
[12:30:16 CET] <thebombzen> hmm I'm having an issue with -vf tile
[12:30:26 CET] <thebombzen> I have nine frames labeled 1.jpg through 9.jpg
[12:30:40 CET] <thebombzen> I'm trying to tile them with ffmpeg -i %d.jpg -vf tile=3x3 out.jpg
[12:30:51 CET] <thebombzen> but it keeps overlaying them all on the upper left corner
[13:17:53 CET] <Pandela> Anyone remember any old ffmpeg versions that had interesting bugs? like video glitches and such
[13:55:01 CET] <k13nox> Hi, i have Webm with: Video/Audio at the beginning, only Audio in the middle and finally Video/Audio at the end of my Webm file. When i try to convert to Mp4, i've duplicate frame in the middle. But i need black frame. How can i do that ?
[15:30:38 CET] <basiclaser> hi
[15:30:57 CET] <basiclaser> i am trying to append the live date to a video, with the following code
[15:31:06 CET] <basiclaser> https://www.irccloud.com/pastebin/q1AH5DXl/
[15:31:39 CET] <basiclaser> however, the date variable is not updated, how can i go about updating the date on every frame? or does drawtext not reevaluate itself?
[15:32:33 CET] <furq> you only set $NOW once
[15:32:39 CET] <furq> also drawtext has a strftime option
[15:33:15 CET] <basiclaser> how would i use that?
[15:33:57 CET] <furq> text=%{localtime\:%c}'
[15:35:04 CET] <basiclaser> im seeing lots of references to time in the docs
[15:35:30 CET] <basiclaser> furq: would i be able to set the time zone? as i'm not anywhere near the same timezone as the thing im recording
[15:35:55 CET] <basiclaser> or perhaps i already did
[15:36:40 CET] <furq> i believe it obeys $TZ if it's set
[15:37:27 CET] <vtj> what is the difference between "width" and "coded_width" in the output of ffprobe?
[15:38:16 CET] <furq> coded_width is the actual width, width is the display width
[15:41:47 CET] <vtj> thanks
[15:43:49 CET] <furq> if it's h264 then h264 is always mod16 internally and has cropping info in the bitstream
[15:44:01 CET] <duvrai> Hi. I'm having a problem with ffmpeg not accepting the timestamps from the v4l2 driver pkt_pts:NOPTS, resulting in bad sync, not sure if the problem is ffmpeg or the source: https://github.com/raspberrypi/linux/issues/1836
[15:44:45 CET] <furq> by "display width" i mean the width after that cropping is applied, not after aspect ratios etc are applied
[16:03:47 CET] <user128> hi, does ffmpeg support decoding amr-wb+ (codec id is "sawp")? I am guessing not. I checked the source-code and there is no mention of "sawp" but I would like to confirm.
[16:04:40 CET] <furq> it decodes amrwb, no idea about amrwb+
[16:10:45 CET] <durandal_1707> user128: have sample to share?
[16:24:23 CET] <duvrai> or is my question about timestamps more an #ffmpeg-devel problem?
[16:33:03 CET] <user128> durandal_1707, I only have some encrypted samples. I am not sure if my decryption program works 100% correctly (ffmpeg quits after seeing the "sawp" codec id). it seems amr-wb+ isn't supported by anything open-source. even the proprietary codec seems hard to obtain.
[16:39:53 CET] <DHE> duvrai: ffmpeg-devel is for people looking to update/patch ffmpeg itself. the main devs are there
[16:53:18 CET] <durandal_1707> user128: you can try settibg decoder manually.
[16:58:57 CET] <user128> durandal_1707, I found a decoder / encoder at https://portal.3gpp.org/desktopmodules/Specifications/SpecificationDetails.aspx?specificationId=1451. however it doesn't decode my decrypted .3gp files which use "sawp" codec.
[17:01:06 CET] <durandal_1707> user128: probably because it expects different layout of data
[17:03:34 CET] <user128> user128, my decrypted data might be wrong (mediainfo does recognize the container format and codec id correctly). playing with the official encode + decoder to better understand how they work.
[17:03:42 CET] <user128> durandal_1707, ^^
[18:30:39 CET] <duvrai> So I'm capturing a v4l2 stream, but the first frame has missing or corrupted timestamp info.
[18:32:25 CET] <duvrai> the bad frame is a h264 keyframe
[18:35:13 CET] <duvrai> when muxing with audio it doesn't sync, and the muxer complains "Timestamps are unset in a packet for stream 0"
[18:36:44 CET] <duvrai> So how do I ignore the first frame for syncing?
[18:48:33 CET] <genii> Hi. How accurately can -itsoffset be set? Obviously within 1/10th with something like -itsoffset 1.1  ...but can it be adjusted within 1/100th or 1/1000th as well?
[18:49:26 CET] <genii> ( for synchronizing stereo high-speed feeds)
[19:19:09 CET] <cfoch-always> hi
[19:19:16 CET] <cfoch-always> how can I screencast in Wayland?
[19:20:34 CET] <vans163> Aynone know where to find the full wire format for SPS frame? I looking here but my SPS frame is longer and different.  https://cardinalpeak.com/blog/the-h-264-sequence-parameter-set/
[19:27:59 CET] <Miyagui> Hi all , i have a question for u
[19:28:52 CET] <Miyagui> i am scaling an a gif with the size of a video and then overlay . And i want to keep the aspect ratio of the Gif
[19:29:04 CET] <jkqxz> vans163:  <https://www.itu.int/rec/T-REC-H.264>, section 7.3.2.1.1.
[19:29:10 CET] <Miyagui> i am using [1][0]scale2ref=iw:ih:force_original_aspect_ratio=decrease[overlay][video]
[19:29:32 CET] <Miyagui> but the force_original_aspect_ratio
[19:29:41 CET] <Miyagui> not working at all
[19:31:27 CET] <vans163> jkqxz: sweet thanks, this looks like my frame now
[19:33:02 CET] <kerio> what does -tune fastdecode do
[19:36:31 CET] <Miyagui> "-tune fastdecode  disables CABAC and the in-loop deblocking filter to allow for faster decoding on devices with lower computational power."
[19:36:45 CET] <kerio> ty
[19:36:49 CET] <kerio> where's that from?
[19:38:03 CET] <Miyagui> http://superuser.com/questions/564402/explanation-of-x264-tune there is more info there
[19:39:03 CET] <relaxed> kerio: also, x264 --fullhelp | less
[19:39:48 CET] <Miyagui> any idea why force_original_aspect_ratio not working in : [1][0]scale2ref=iw:ih:force_original_aspect_ratio=decrease[overlay][video]
[19:39:50 CET] <Miyagui> ?
[19:42:34 CET] <vans163> Anyone aware how to parse this ue(v)  https://en.wikipedia.org/wiki/Exponential-Golomb_coding   makes no sense
[19:42:50 CET] <vans163> right after level_idc u(8)    seq_parameter_set_id 0 ue(v)
[19:43:17 CET] <vans163> does that mean ignore / skip 0 bytes, and continue parsing
[19:43:42 CET] <vans163> basically saying the bytes that follow can be variable size due to the if conditions?
[19:46:25 CET] <jkqxz> vans163:  The exp-golomb decoding is defined in section 9.1.
[19:48:48 CET] <vans163> jkqxz: thanks looking
[20:52:53 CET] <Miyagui> When you use [1][0]scale2ref  the input is the second parameter
[20:53:13 CET] <Miyagui> Is there any way to make reference to the fist one?
[21:30:37 CET] <rdz_> it there a way to preload a network input (RTMP or HLS ) ?
[21:31:15 CET] <rdz_> I'm getting same lag without errors on the ffmpeg output and I'm gessing it is network related
[21:31:39 CET] <rdz_> demuxpreload doesn't work with HTTP input
[21:48:19 CET] <nas> I'm unable to register to trac because it says my ip is SPAM
[21:51:23 CET] <durandal_1707> use vpn
[21:51:33 CET] <durandal_1707> or proxy
[22:10:51 CET] <nas> ok i do so but i get no mail. And now the proxy is also spam.
[22:33:35 CET] <durandal_1707> chose another proxy, repeat
[22:41:46 CET] <baldgeek_> hi
[22:48:38 CET] <durandal_1707> hi
[22:49:13 CET] <baldgeek_> So I'm trying to transcode a h.265 stream to h.264 in a mpeg_ts containter, and it works when I transcode to a file but crashes when I transcode to a stream that I want to server with ffserver
[22:49:14 CET] <baldgeek_> http://pastebin.com/PLB1jE8s
[22:50:07 CET] <BtbN> You'll have a hard time finding help or anyone that cares about ffserver.
[22:50:34 CET] <baldgeek_> any other open servers I could use?
[22:53:06 CET] <baldgeek_> I'm in the R&D portion of this task, so I'm open to something else at this point :)
[22:56:34 CET] <BtbN> well, by now nginx-rtmp is the only thing that somewhat works
[22:56:41 CET] <BtbN> or just ffmpeg hls output + any http server
[23:01:10 CET] <baldgeek_> hmmm
[23:01:40 CET] <baldgeek_> ok.. I'll look into that. honestly, treading water as this point and learning as I go
[23:22:29 CET] <vans163> anyone know the use for framerate density?
[23:22:55 CET] <vans163> why shuld it ever be higher then 1 for an encoder?
[00:00:00 CET] --- Sat Feb 11 2017