[Ffmpeg-devel-irc] ffmpeg.log.20200128

[burek]

[01:32:39 CET] <danilo82> hello, i want to remove Stream #0:1, Stream #0:2, Stream #0:5, Stream #0:6 from a video
[01:42:14 CET] <relaxed> danilo82: ffmpeg -i input -map 0:0 -map 0:3 -map 0:4 -c copy output
[02:05:04 CET] <Conjecture> Is flac not entirely lossless?
[02:05:30 CET] <Conjecture> I was told there was something to do with padding that it did that made it such that it wasn't completely lossless
[02:06:30 CET] <c_14> iirc it pads 16bit to 24bit if that's what you mean
[02:09:24 CET] <c_14> I think that's just me misremembering shit though
[02:11:52 CET] <c_14> you can always test any file you encode by decoding the flac and comparing to the raw input stream
[02:18:11 CET] <furq> what does "not entirely lossless" mean
[02:18:19 CET] <furq> it's either lossless or it isn't
[02:23:20 CET] <pink_mist> well it doesn't support floating point samples, only integer samples, and no more than 32 bits per sample, and no more than 8 audio channels ... at least according to wikipedia :P
[02:44:07 CET] <c3r1c3-Win> Just padding from 16 to 24 bits would still be lossless.
[02:44:33 CET] <cehoyos> pink_mist: More than 32bits per sample are unlikely
[02:44:44 CET] <pink_mist> yeah, that's true
[02:54:36 CET] <furq> it doesn't support 32-bit int either
[02:54:40 CET] <furq> if anyone actually uses that
[02:55:10 CET] <furq> technically it does but practically it doesn't
[02:57:09 CET] <furq> also none of that makes flac not lossless, it just means the preprocessing step might not be lossless
[03:01:37 CET] <klaxa> just like gif, lol
[03:05:45 CET] <linext> can anyone help me add titles to a .xspf playlist?
[03:42:14 CET] <cehoyos> furq: Even FFmpeg's flac encoder supports 32bit integer
[03:49:27 CET] <ponyrider> linext: does ffmpeg even do that?
[03:49:40 CET] <linext> ponyrider, i figured it out
[03:49:59 CET] <linext> no, i asked in VLC but no one is available to responde
[03:50:04 CET] <furq> cehoyos: how do you get that to work
[03:50:14 CET] <furq> if you pass it s32 it'll encode 24-bit
[03:50:51 CET] <linext> https://i.imgur.com/8TDVyzR.png
[03:51:09 CET] <furq> also last i checked libflac won't encode or decode it, so it's not really usable
[03:51:28 CET] <linext> i've got an hdhomerun with a quad tuner
[03:51:43 CET] <linext> ffmpeg does a good job recording tv like a DVR
[03:52:41 CET] <cehoyos> furq: Sorry, I didn't check carefully enought
[03:53:09 CET] <linext> i'm considering write a web app to schedule when ffmpeg records
[03:53:18 CET] <furq> there is definitely an encoder that does it and iirc ffmpeg's flac decoder supports it
[03:53:35 CET] <furq> maybe flake
[03:54:36 CET] <furq> but if you actually need s32 for some perverse reason then you should probably just use wavpack
[03:55:15 CET] <furq> or just use fltp wavpack
[07:24:09 CET] <Conjecture> Thanks for the help with my flac lossless question
[07:24:18 CET] <Conjecture> Good night
[11:37:58 CET] <snooky> hi all
[11:41:45 CET] <snooky> Hi there. I have a folder with subfolders on my nas. in this are films and tv shows. now i want to make a script (bash) which lists all files, searches, and brings all series to X MB file size and all films to YMB file size. so that in the end all films and tv shows have the same files. I need twopass for that, right?
[11:42:58 CET] <snooky> then everyone knows the problem. effects. music. sounds and so on is loud. and the voices of the film quietly. sound effects extremely loud again. is there a way to bring the ffmpeg the audio files to the exact same volume? So that an explosion is as loud as the voices?
[11:43:44 CET] <funnybunny2> Why is there an av_strerror and an av_make_error_string? The latter just calls the former.
[11:44:20 CET] <funnybunny2> I guess legacy reasons? Which one should I use?
[11:47:12 CET] <relaxed> snooky: copy the audio stream
[11:47:43 CET] <snooky> and than?
[11:48:39 CET] <snooky> i want a constant audio volume
[11:48:57 CET] <relaxed> well, I wouldn't transcode my entire library to save space on a NAS
[11:50:35 CET] <snooky> I have enough space. it's not about space. audio it annoys me that the effects are louder and the rest is quiet. I want the volume to be the same. ffmpeg normalize?
[11:51:10 CET] <relaxed> look at the dynaudnorm filter
[11:51:41 CET] <snooky> it's about. my NAS in online. so that I can access it wherever I am via vpn. since I travel a lot, that's practical. but it buffers too much when I have a film of 8/9 gb .. so it has to go down to 3 / 4gb so that the stream runs without buffers.
[11:54:10 CET] <funnybunny2> Is there a way to know in advance how big of a buffer I need to allocate for my PCM samples?
[11:55:48 CET] <funnybunny2> av_samples_get_buffer_size I guess
[12:20:02 CET] <funnybunny2> av_samples_get_buffer_size is giving me a way-too small number compared to how many bytes I seem to be actually receiving
[12:23:23 CET] <funnybunny2> I have a 1184-byte 2-channel .wav file which I'm decoding to s16 pcm. av_samples_get_buffer_size returns 192000, but if I do count += frame->linesize[0] * avcodec_context->channels; for each frame, I get a total of 2421760
[12:26:31 CET] <funnybunny2> Ah, I think I see what I messed up...
[12:34:09 CET] <funnybunny2> OK
[12:34:46 CET] <funnybunny2> I'm not sure how I'm supposed to know what to pass for nb_samples to av_samples_get_buffer_size. It says "the number of samples in a single channel"
[12:36:46 CET] <funnybunny2> It would seem that I need the audio duration...
[12:42:30 CET] <funnybunny2> Where is everyone?
[12:52:02 CET] <JEEB> funnybunny2: $dayjob
[12:52:38 CET] <DHE> asleep
[13:11:18 CET] <funnybunny2> I'm in the wrong country
[13:31:05 CET] <ha3mak> hi
[13:31:16 CET] <ha3mak> I'm trying to create a HLS stream from my IP cam with ffmpeg
[13:31:25 CET] <ha3mak> I'm doing it like this: ffmpeg -i rtsp://username:password@mycam.tld/videoMain -c:v libx264 -crf 21 -preset veryfast -g 25 -keyint_min 25 -sc_threshold 0 -vsync 2 -an -f hls -hls_time 6 -hls_list_size 3 -hls_flags delete_segments -hls_playlist_type event /tmp/hls/stream.m3u8
[13:31:46 CET] <ha3mak> It works but it never deletes a segment
[13:32:08 CET] <ha3mak> It just appends the newer ones to the m3u8 and also keeps them on the disk
[13:32:33 CET] <ha3mak> I'm running ffmpeg 4.2.2, I think it's the latest
[13:32:59 CET] <ha3mak> Is it a bug or I'm doing someting bad?
[14:01:14 CET] <ha3mak> Hmm, I found out if I don't use the -hls_playlist_type event option it works
[14:01:59 CET] <BtbN> Sounds about right, since you have a live playlist from how you describe it
[14:02:19 CET] <BtbN> An event playlist does not change, and only adds elements to the end
[14:03:04 CET] <ha3mak> Thanks, I didn't knew it. My bad, sry :)
[17:32:35 CET] <snooky> https://nopaste.xyz/?2c2f0b31a3b4c9de#VII7xFGCY+48jbAC/ngNGdGPWY9YpX9e9czV8M2K/d0=
[17:32:36 CET] <snooky> hi all
[17:33:01 CET] <snooky> https://nopaste.xyz/?061c33d1fc02a5f5#VttY0rS+YYLU1HF0Ou+5LIX98nOIisPswIpkAL1vhmw=
[17:33:04 CET] <snooky> my script
[17:33:07 CET] <snooky> what is wrong!?
[17:34:07 CET] <BtbN> Well, what is it supposed to do?
[17:34:30 CET] <BtbN> Immediate thing that springs to eye is that it looks like you are trying to do two pass encoding, while not using an encoder in the first pass.
[17:35:48 CET] <snooky> oh yes
[17:35:49 CET] <snooky> thanks
[17:35:49 CET] <TanaPanda> why did you name yourself snooky
[17:35:51 CET] <snooky> now i runs
[17:36:50 CET] <snooky> ok the encodeing runs
[17:37:04 CET] <snooky> now i want an strict volume level
[17:37:50 CET] <snooky> so if I completely re-encode the film anyway. can I then immediately edit the sound so that the level is completely the same? So the effects are as loud as they are true?
[17:43:14 CET] <BtbN> loudnorm also has a two-pass mode you can use
[17:48:11 CET] <matt1122> Hello, I have live RTMP input (from red5pro) from iPhone that's either 1280x720 OR 720x1280 ... and it will constantly be changing between ... is it going to be possible to handle this w/ ffmpeg so that the output resolution changes on the fly?
[17:49:19 CET] <JEEB> at least the API has no limitations with that. in the worst case you might need to reconfig your encoder, although various encoder wrappers might already be handling that in the function that receives your AVFrame
[17:59:33 CET] <matt1122> Thx JEEB
[18:36:51 CET] <matt1122> JEEB Is there a way to handle the resolution change (phone rotation) on the fly when using the FFMPEG command?
[18:38:29 CET] <matt1122> Currently, when I rotate the phone the aspect ratio stays at whatever the initial aspect ratio was.
[19:17:12 CET] <linext> anyone know what GOP stands for?  https://imgur.com/a/1mXvhwG
[19:17:16 CET] <linext> is it group of pictures?
[19:24:43 CET] <matt1122> I think that it's GOP=group of pictures
[19:30:27 CET] <linext> do you know what it does?
[19:30:28 CET] <Mavrik> linext, yes, the interval between keyframes (roughly)
[19:30:40 CET] <linext> ah, so 1 keyframe for 30 in my screenshot?
[19:30:46 CET] <Mavrik> Yeah
[19:30:55 CET] <Mavrik> Depending on encoder it might not be static
[19:31:02 CET] <Mavrik> E.g. it might insert it sooner but never later.
[19:31:35 CET] <linext> i plugged a chromecast into a ISV-HEM4
[19:31:43 CET] <linext> the picture is good at 720p
[19:32:01 CET] <linext> no PDF manual or support
[21:50:01 CET] <plitter> ffmpeg -video_size 1920x1080 -framerate 30 -f x11grab -i :0.0+0,0 -f pulse -i default -f v4l2 -input_format mjpeg -framerate 30 -video_size 640x480 -i /dev/video0 -map 0 -map 1 -map 2 output.mkv I'm having this command get mic, monitor and desktop, but the sound is off from the cam... I thought since I specified all the relevant input to start at the same time that it would also be synchronized...
[21:50:02 CET] <plitter> is there something I'm doing wrong or is there something else that I have to do?
[22:10:13 CET] <plitter> it also has a tendency to stop recording audio before the video stops...
[22:32:21 CET] <cehoyos> plitter: Please provide the command line you tested together with the complete, uncut console output
[22:50:48 CET] <plitter> cehoyos: https://clbin.com/kNbjZ
[22:52:44 CET] <cehoyos> Try with preset ultrafast, your hardware seems too slow
[22:53:57 CET] <cehoyos> And / or remove the map options, you seem to encode video twice
[22:54:35 CET] <cehoyos> You want to encode two video streams?
[22:54:45 CET] <cehoyos> Then it is normal that the hardware can't do it
[22:55:25 CET] <cehoyos> And consider using -pix_fmt yuv420p (x2) if you want the output file to be playable on non-FFmpeg based players
[22:58:19 CET] <plitter> my thought is to get all streams together to later put together into a video for youtube or similar, basically I want a face cam to be put on top of my screen recording with audio
[23:02:04 CET] <plitter> cehoyos: https://clbin.com/m7x2m
[23:02:12 CET] <plitter> I forgot to show my command...
[23:02:53 CET] <cehoyos> speed=1.08x indicates that AV sync can be reached (after some time)
[23:03:16 CET] <cehoyos> Sorry:
[23:03:25 CET] <plitter> so I should let it run longer?
[23:03:39 CET] <cehoyos> The third input has an incorrect start time 108690.846475 that makes synchronisation impossible
[23:04:33 CET] <plitter> I saw some weird time stamps when I was testing with mpv as well...
[23:04:37 CET] <cehoyos> There is an v4l2 input option -timestamps abs that you should test
[23:05:26 CET] <plitter> ok, give me some minutes :)
[23:17:37 CET] <plitter> I tried with -timestamps now but it fails, saying "Error setting option timestamps to value now"
[23:39:50 CET] <cehoyos> I didn't suggest "now" as value for the option -timestamps...
[23:44:19 CET] <plitter> I thought I was supposed to set my own time....
[23:47:15 CET] <plitter> https://clbin.com/RaWrF when using abs the audio is still coming after the video
[00:00:00 CET] --- Wed Jan 29 2020