[Ffmpeg-devel-irc] ffmpeg.log.20191101

[burek]

[01:08:41 CET] <KombuchaKip> JEEB: Alright, so found the problem. Not a damn thing wrong with ffmpeg. I was just doing some pointer arithmetic wrong. Thanks for your help.
[01:58:23 CET] <a0z> I'm trying to rip subs from a few VOB files, but the subtitle streams doesn't start until 25megs into the VOB file, how do I use ffmpeg to start the stream processing from the starting point of the subs stream? see here for what i tried: http://paste.debian.net/1112440/  interesting around lines 57-64
[01:59:20 CET] <pink_mist> why does it matter if it starts 25MB in?
[02:00:08 CET] <a0z> well, when i try to rip the subs ffmpeg says there is no subtitle stream, however when transcoding the audio/video there was no problem as ffmpeg finds those streams okay
[02:02:58 CET] <nicolas17> a0z: note that DVD subtitles are image overlays, so you can't get them into srt (text) format... unless you do character recognition
[02:03:09 CET] <a0z> ah, i see
[02:03:38 CET] <nicolas17> but that's not an answer to your problem :)
[02:04:41 CET] <furq> a0z: -probesize 30M
[02:05:05 CET] <nicolas17> furq: that goes before -i right?
[02:05:10 CET] <furq> yeah
[02:05:21 CET] <furq> you might also need to set -analyzeduration to some value higher than 5M
[02:31:35 CET] <a0z> thanks, that did it!
[06:26:42 CET] <pk08> hi, is there any way to find muxing mode of closed caption from mpeg-ts?
[06:27:15 CET] <pk08> in ffprobe it shows "ATSC A53 Part 4 Closed Captions" for every CC types
[06:27:42 CET] <pk08> but i need to know if its SCTE 128 or A53
[10:13:23 CET] <jancoow8> Hi, this should cut a video from 06:39:00 to 07:39:30 right
[10:13:24 CET] <jancoow8> ffmpeg -ss 06:39:00 -i ThunderDome\ 2019\ Full\ Version\!-YtxMfkRjhjM.mkv -codec copy -t 07:39:30 sefa.mkv
[10:13:43 CET] <JEEB> t is duration
[10:13:59 CET] <jancoow8> Ah right, that's why it isn't working
[10:14:00 CET] <JEEB> so that cuts from the closest random access point before 06:39:00
[10:14:10 CET] <JEEB> (or at that point, if it has a random access point at that point)
[10:14:30 CET] <JEEB> and then cuts off a thing with that duration
[10:14:37 CET] <jancoow8> so how can I set an end time
[10:15:06 CET] <JEEB> I think there might be something like -to
[10:15:39 CET] <JEEB> -to time_stop       record or transcode stop time
[10:15:52 CET] <jancoow8> Ah right!
[10:15:53 CET] <jancoow8> Thanks!
[10:35:56 CET] <Botik2> hello
[10:36:24 CET] <Botik2> I need help with retrieving 1 second interval frames from 29.97 fps video. Here is what I have tried:
[10:36:38 CET] <Botik2> ffmpeg -i rugby\ football_1013.mp4 -vf "select='not(mod(n,30))',setpts='N/(30*TB)-(3*round(N/3000))/(30*TB)'" -f image2 rugby%08d.jpg
[10:37:49 CET] <Botik2> The idea was to just retrieve every 30th frame and shift time by 3 frames every 3000 frames to compensate for difference between 29.97 and 30 fps. But it is not giving me desired results
[10:40:22 CET] <Botik2> but i am getting 15 second shift around 84 minute. What am I doing wrong?
[12:18:43 CET] <CounterPillow> JEEB: doesn't -ss before -i break -to?
[12:19:38 CET] <CounterPillow> as in, -to starts behaving like -t because fast seek sets the time base to 0
[12:22:09 CET] <JEEB> CounterPillow: no idea tbqh :P
[12:22:16 CET] <JEEB> I prefer -t since it's specific
[12:22:41 CET] <CounterPillow> I just do the math myself, but it would be really nice if that thing were fixed
[12:30:42 CET] <CounterPillow> Though to fix this we probably first need to look into bofa
[12:39:44 CET] <Adcock> when will rav1e
[12:39:59 CET] <Adcock> be usable using ffmpeg
[12:40:01 CET] <Adcock> ?
[12:43:13 CET] <JEEB> as soon as rav1e makes a release so that if they make changes to the C API we can check for the version
[12:43:27 CET] <JEEB> I think the person who posted the patch for rav1e is one of the rav1e devs
[12:43:52 CET] <JEEB> currently they're not doing semantic versioning so there's no way to check :P
[12:44:06 CET] <Adcock> Okay.
[12:44:17 CET] <Adcock> Thanks.
[12:44:18 CET] <JEEB> in other words, until rav1e makes a release or at least starts using semantic versioning, the patch is not going in
[12:44:24 CET] <JEEB> unfortunately
[12:45:02 CET] <Adcock> Do you think rav1e will be a big impact in terms of encoding for normal people?
[12:45:21 CET] <Adcock> Like me who has no dedicated hardware of encoding?
[12:46:47 CET] <Adcock> *for
[13:00:12 CET] <Adcock> Taking it to  #daala.....:D
[13:07:13 CET] <void09> anyone got a oneliner to convert a Bluray folder to .iso (properly) ?
[13:07:20 CET] <void09> on linux
[13:13:58 CET] <AiNA_TE> mkisofs -o blah.iso /home/user/dir_you_wanna_isoify
[13:15:52 CET] <AiNA_TE> man mkisofs
[13:18:36 CET] <void09> it's not that simple, filesystem has to be udf.. some version of it
[13:23:20 CET] <AiNA_TE> google-fu says
[13:23:23 CET] <AiNA_TE> mkisofs -udf -allow-limited-size -input-charset "utf-8" -o "$TargetFile" "$SourceDir"
[13:23:34 CET] <AiNA_TE> but i cannot verify it, give it a shot though
[13:31:02 CET] <void09> needs to be version 2.50
[15:53:16 CET] <Adcock> Guys how do I know if my GPU or CPU has hardware decoding for x264 or x265?
[15:53:38 CET] <Adcock> Any way to test it?
[15:58:46 CET] <DHE> you still need drivers, so you'll have to consult those drivers
[15:59:07 CET] <DHE> on linux I'd run nvidia-settings and check what its capabilities are for vdpau
[15:59:44 CET] <Mavrik> IIRC ffmpeg can list that too
[15:59:49 CET] <Mavrik> Depending on OS
[16:18:11 CET] <void09> how to output the resolution of a png image ?
[16:18:48 CET] <void09> ffprobe.. ok
[16:19:04 CET] <furq> ffprobe -v error -select_streams v:0 -show_entries stream=width,height -of csv=nk=1:p=0 foo.png | tr ',' 'x'
[16:19:46 CET] <void09> nice thanks furq
[16:23:20 CET] <relaxed> xsv
[16:28:58 CET] <Radioactive> heyyyy
[16:29:01 CET] <Radioactive> [ac3 @ 00000183297DE4C0] Specified channel layout '7.1' is not supported
[16:29:16 CET] <Radioactive> since when was that a thing? ac3 used to support 7.1 just fine
[16:30:07 CET] <furq> you're thinking of eac3/dd+
[16:30:37 CET] <Radioactive> this is for encoding
[16:31:10 CET] <Radioactive> with libavcodec
[16:31:28 CET] <Radioactive> basically i used to be able to select AV_CODEC_ID_AC3
[16:31:45 CET] <JEEB> changes related to that seem to be from like 2010
[16:31:58 CET] <Radioactive> is there any way to export ac3 with 8 channels?
[16:32:02 CET] <Adcock> Mavrik: How?
[16:32:07 CET] <JEEB> I would have been surprised if it ever worked
[16:32:25 CET] <Radioactive> before i would do something like this _codecContext->channel_layout = av_get_default_channel_layout(numChans);
[16:32:27 CET] <Radioactive> and it would work fine
[16:32:37 CET] <Radioactive> with ac3, and mp2 as well
[16:32:41 CET] <Radioactive> now both give that error
[16:32:48 CET] <Radioactive> where numChans could be 8
[16:34:21 CET] <JEEB> 1aeb88b77ddbd3802127a61bf8610e55b6c49a6f from 2011 added the check for both AC3 encoders
[16:34:27 CET] <JEEB> the integer only and the float one
[16:35:51 CET] <JEEB> although that just renames the list it seems
[16:35:59 CET] <JEEB> the list itself being there is even older
[16:37:03 CET] <Radioactive> what does that mean "the check"?
[16:38:05 CET] <JEEB> added a .channel_layouts attribute to the AC-3 encoder AVCodecs. although, as I noted I then checked the commit and the change was actually just renaming the lists
[16:38:09 CET] <JEEB> not adding them there
[16:38:34 CET] <Radioactive> is there somewhere i can see a list of channel layouts supported for ac3? (and mp2)
[16:38:51 CET] <JEEB> b33dae5efbc23cafaeb646231a949c296574540b
[16:38:59 CET] <JEEB> that is where the list seems to have gotten added
[16:39:21 CET] <JEEB> Radioactive: yes, definitely
[16:39:25 CET] <JEEB> since ffmpeg.c can also list them
[16:39:27 CET] <furq> Radioactive: -h encoder=ac3
[16:39:37 CET] <JEEB> ffmpeg -h encoder=ac3 will output them, and they are available through the API
[16:39:56 CET] <Radioactive> hmm cool
[16:41:01 CET] <JEEB> yea, so it loops through the array of channel_layouts in the AVCodec
[16:41:37 CET] <JEEB> so if you have a AVCodec *codec , you can if (codec->channel_layouts) { for (...) } 
[16:44:53 CET] <JEEB> (such lists should end with 0/NULL entry at the end)
[16:45:07 CET] <Radioactive> mp2: "Supported channel layouts: mono stereo"
[16:45:09 CET] <Radioactive> whaaaaaaa
[16:45:18 CET] <Radioactive> i could have sworn mp2 supported multichannel audio
[16:45:25 CET] <Radioactive> am i going crazy?
[16:45:53 CET] <JEEB> yea I don't remember it ever doing that
[16:46:04 CET] <Radioactive> maybe i didn't define an explicit channel layout before?
[16:46:27 CET] <Radioactive> if i just set the number of channels but not the channel layout
[16:46:34 CET] <Radioactive> and encode that, what happens
[16:46:39 CET] <JEEB> maybe, but I have no idea if that would actually do
[16:46:42 CET] <JEEB> or if it would let you pass
[16:49:44 CET] <Radioactive> only one way to find out
[17:02:10 CET] <Radioactive> "[mp2 @ 0000023219E34150] encoding 8 channel(s) is not allowed in mp2"
[17:02:19 CET] <Radioactive> idk i feel like i've entered an alternate timeline
[17:02:22 CET] <Radioactive> i swear this used to work
[17:04:30 CET] <durandal_1707> never worked
[17:06:16 CET] <DHE> maybe you encoded to a different format?
[17:12:52 CET] <FOSSie_bear> hi, I have been searching for some solution to my problems when I try to convert a black white video to h264. Unfortunatley the picture gets a "greenish" tint all the time. I guess it is due to color space conversion...
[17:13:27 CET] <FOSSie_bear> Has anyone had these problems before? Can you recommend a guide to correct these issues?
[17:16:15 CET] <durandal_1707> full command and uncut console output missing
[17:17:02 CET] <FOSSie_bear> Actually the video input is "ffv1 with bgr0(progressive)" and the output is "h264_nvenc with bgr0". Perhaps I quickly try libx264 first and then post the outputs
[17:22:39 CET] <FOSSie_bear> https://pastebin.com/rcnY83Ct
[17:25:35 CET] <durandal_1707> i do not think that is ffmpeg bug, both input and output are bgr0
[17:28:57 CET] <FOSSie_bear> ok. I was not sure whether h264_nvec has compatibility issues with bgr0
[17:29:32 CET] <FOSSie_bear> perhaps there are some internal conversions before hardware upload and after download :/
[19:09:41 CET] <realies> would it possible to use something like this as a ffmpeg filter? https://github.com/xiph/rnnoise/blob/master/examples/rnnoise_demo.c
[19:11:35 CET] <durandal_1707> realies: already done
[19:12:35 CET] <durandal_1707> http://ffmpeg.org/ffmpeg-filters.html#arnndn
[19:13:26 CET] <realies> lol, guess I just need to feed it the model?
[19:13:39 CET] <durandal_1707> yes
[19:13:53 CET] <durandal_1707> see relevant github repo
[19:15:25 CET] <realies> not sure were that is in xiph's rnnoise
[19:15:40 CET] <realies> or where you could find trained models?
[19:15:46 CET] <durandal_1707> another repo
[19:15:54 CET] <durandal_1707> use search
[19:16:23 CET] <durandal_1707> rnnoise-models
[19:17:41 CET] <realies> :o
[20:35:48 CET] <Retal> hello guys, i have many errors: Error while decoding stream #0:1: Error number -16976906 occurred
[20:36:20 CET] <Retal> what means error 16976906?
[21:04:22 CET] <realies> seems like this filter chain does not apply the noise reduction 'dynaudnorm=b=true:m=100:c=true, arnndn=model=/Users/realies/Documents/Projects/rnnoise-models/beguiling-drafter-2018-08-30/bd.rnnn, compand=0:1:-90/-900|-70/-70|-30/-9|0/-3:6:0:0:0,
[21:04:22 CET] <realies> loudnorm=i=-19:tp=-1:lra=6:measured_I=-25.97:measured_LRA=6.70:measured_TP=-3.34:measured_thresh=-36.40:offset=0.05:linear=true:print_format=json'
[21:04:24 CET] <realies> any ideas why?
[21:05:05 CET] <realies> separating it in two forks of ffmpeg seems to do it though
[21:11:06 CET] <realies> weird
[21:17:08 CET] <realies> nah, running in two forks does nothing either
[21:23:51 CET] <durandal_1707> realies: whatever you do, you do nonsense filtering
[21:23:59 CET] <realies> how come?
[21:24:59 CET] <durandal_1707> you can not apply arnndn after dynaudnorm
[21:25:14 CET] <realies> I thought the filters process the input one after another
[21:27:06 CET] <realies> aaand wrong paths, seems to be working ok
[21:28:07 CET] <realies> durandal_1707, why not arnndn after dynaudnorm though?
[21:28:37 CET] <durandal_1707> because it changes audio volume
[21:30:18 CET] <realies> arnndn seems to be able to handle it well
[21:30:28 CET] <realies> at least bd.rnnn
[21:31:03 CET] <realies> gotta do some A/B comparisons, though that by exaggerating the noise floor it might perform better suppressing the noise
[23:39:41 CET] <Retal> i have many errors: Error while decoding stream #0:1: Error number -16976906 occurred What means error 16976906?
[23:40:34 CET] <JEEB> try av_err2str ?
[23:40:51 CET] <JEEB> unless that's what ffmpeg.c outputs and you're just looking at its terminal output
[23:41:16 CET] <pink_mist> it certainly looks like terminal output to my inexperienced eyes
[23:41:36 CET] <pink_mist> why are error numbers in the negative though? that seems very odd as well
[23:42:49 CET] <klaxa> because positive numbers could be valid
[23:43:09 CET] <klaxa> for whatever function returns the error
[23:44:38 CET] <pink_mist> ah, I guess I see the point of it then
[00:00:00 CET] --- Sat Nov  2 2019