[Ffmpeg-devel-irc] ffmpeg.log.20151028

burek burek021 at gmail.com
Thu Oct 29 02:05:01 CET 2015


[00:02:33 CET] <Mavrik> constrained baseline is nowhere near your issue.
[00:03:52 CET] <jgarde> Then I'm at a total loss. I can get the stream published onto Periscope's RTMP servers just fine and playback using the online client, but I can't get their Periscope client to do anything but crash.
[00:15:46 CET] <Lectem> Hi, is there some place where I can see the dependancies between demuxers ?
[00:16:20 CET] <fred1807> can I command ffmpeg to slice the video in X equal parts (without the need to specify duration/seconds) ?
[01:20:00 CET] <Lectem> ... is the mailing list password stored in cleartext?
[01:20:14 CET] <Lectem> just received a mail with my password in it
[01:20:47 CET] <rsully> I always imagined all mailinglist software being written in the 90s
[01:20:49 CET] <rsully> so probably
[01:21:26 CET] <Lectem> ...
[01:21:51 CET] <rsully> not that i actually have any factual input to add to this
[01:25:21 CET] <Lectem> http://www.gnu.org/software/mailman/faq.html The password used to do the initial authentication for any non-public Mailman page is sent as clear text over the net.
[01:25:31 CET] Action: Lectem facepalms
[04:54:15 CET] <jasom> Lectem: IIRC mailman's account creation page specifically tells you to *not* use a high value password for that reason
[04:59:14 CET] <jgarde> Nothing but the best military grade cyphers.
[08:05:36 CET] <foreverska> is there a decent ffmpeg gui?
[08:05:49 CET] <foreverska> (deiban)
[08:42:40 CET] <JEEB> foreverska: no. it can do so many things that abstracting it well is really hard. if you just want a re-encoding gui using the libraries then probably handbrake ia the least awful alternative
[09:06:30 CET] <foreverska> Yea I figured as much.  I just need a start and stop button for a particular script.  Then again "press Q to stop it" would be a 2 second explanation if people weren't so squeamish about terminal outputs.
[13:27:06 CET] <arpu> hello is it possible to generate a 360 degree video with ffmpeg ?
[17:57:16 CET] <Prelude_Zzzzz> hey everyone.. good day.. can i please get some help. http://pastebin.com/zKt4CACd << ===== I am using this code to generate 3 outputs from a single input. I can't seem to get it to outout 2 to NVENC and the other to THE CPU any way to do it ? can someoen help me with it ?
[17:58:19 CET] <DHE> show us the errors you get
[17:59:15 CET] <Prelude_Zzzzz> [nvenc @ 0x27444e0] Preset "ultrafast" is unknown! Supported presets: hp, hq, bd, ll, llhp, llhq, default
[17:59:25 CET] <Prelude_Zzzzz> its because that last param is different
[17:59:53 CET] <DHE> "h264" isn't a particular codec implementation, it's a codec. ffmpeg chose nvenc to do the last job as well. explicitly specify libx264
[17:59:53 CET] <Prelude_Zzzzz> i want to be able to offload from the input 2 to nvenc and 1 to CPU
[18:00:32 CET] <Prelude_Zzzzz> Unknown encoder 'libx264'
[18:00:51 CET] <Prelude_Zzzzz> possibly because i enabled nvenc on the compile
[18:00:59 CET] <DHE> but didn't enable libx264
[18:01:06 CET] <DHE> you can have both. I've done exactly this before
[18:01:07 CET] <Prelude_Zzzzz> let me see.
[18:01:15 CET] <Prelude_Zzzzz> ok one sec let me recompile to see
[18:01:17 CET] <Prelude_Zzzzz> thank you
[18:02:09 CET] <Prelude_Zzzzz> maybe that is my mistake.. and it would be cool if that was the easy part :P
[18:05:43 CET] <Prelude_Zzzzz> OMG .. thank you
[18:05:50 CET] <Prelude_Zzzzz> silly little thing
[18:06:15 CET] <flux> btw, the AVCodec struct has a field for hardware accelerator, and the comment says it's not used for encoders
[18:06:29 CET] <flux> does this hold true for nvenc as well?
[18:06:46 CET] <Prelude_Zzzzz> no idea what it is
[18:07:15 CET] <JEEB> flux: hwaccels are special snowflakes
[18:07:30 CET] <JEEB> aka you have to do work as well
[18:07:53 CET] <JEEB> the nvenc encoder just uses provided nvidia libraries
[18:08:00 CET] <Prelude_Zzzzz> right
[18:08:08 CET] <Prelude_Zzzzz> and it has a built in hw component for encoding
[18:08:14 CET] <Prelude_Zzzzz> there is decoding and encoding
[18:08:18 CET] <flux> ok, so the field was "hwaccel_context" and it's not related to mostly anything, thanks :)
[18:08:52 CET] <flux> it says the vaapi context makes use of it
[18:09:12 CET] <JEEB> i think there's a decoder or two in ffmpeg (cli) that use hwaccels
[18:09:13 CET] <Prelude_Zzzzz> ahh ic.. in the old days it uses to use the software and maybe it used it.. but the new nvidia cards have built in hardware encoding
[18:09:27 CET] <Prelude_Zzzzz> they seperated it.. would be awsome cool if now i can also use the software side like nvcuda :P
[18:09:35 CET] <Prelude_Zzzzz> tons of processing power not in use
[18:10:26 CET] <flux> I guess the most generally applicaple hardware acceleration would be for color space conversions, and that's probably dealt efficiently already.. ?
[18:16:10 CET] <Prelude_Zzzzz> hey, ffmpeg supports the nvenc and does it support the ATI cards yet ?
[18:16:20 CET] <Prelude_Zzzzz> i hear they are faster
[18:16:32 CET] <Prelude_Zzzzz> and don't have the silly limitations of 2 encodings per card
[18:17:30 CET] <BtbN> It's not a silly limitation. It's a hard-wired encoder. That's how many encoding-circuits there are.
[18:20:41 CET] <Prelude_Zzzzz> any way around it ?
[18:20:52 CET] <Prelude_Zzzzz> they want you to upgrade to more expensive cards to do it
[18:21:35 CET] <BtbN> Only the GRID Cards have more than 2 encoding units.
[18:21:48 CET] <BtbN> And those are designed for their Game-Streaming-Server-Boxes.
[18:22:16 CET] <Prelude_Zzzzz> i am using the k4200 right now
[18:22:30 CET] <BtbN> That's just a Quadro.
[18:22:34 CET] <Prelude_Zzzzz> but has some fps limitation of like 65fps per instance of encoding on hq
[18:22:53 CET] <Prelude_Zzzzz> ya t he 4200 has no limitation. i am currnly encoding 15 streams
[18:22:54 CET] <BtbN> My GT640 does 1000 FPS+ on HQ.
[18:24:00 CET] <Prelude_Zzzzz> it is strange actually why it colapses .. that script i am running   ( http://pastebin.com/zKt4CACd ) .. if i put that last param bit rate on the GPU it starts to drop fps on everything
[18:24:12 CET] <Prelude_Zzzzz> i can't seem to be stable at that.. so i have to offload  one of them to CPU and then it works
[18:24:26 CET] <Prelude_Zzzzz> but.. i can have like 10 channels running with 2 bit rates each and they all do 30fps
[18:24:39 CET] <Prelude_Zzzzz> but the min i try and increase it to 90fps from one instance it acts up.. not sure why
[18:29:59 CET] <Prelude_Zzzzz> http://developer.download.nvidia.com/compute/nvenc/v4.0/NVENC_AppNote.pdf
[18:32:20 CET] <BtbN> That's outdated.
[18:35:16 CET] <Prelude_Zzzzz> hey what is the real difference between llhq and just hq
[18:35:22 CET] <Prelude_Zzzzz> trying to understand taht
[18:35:31 CET] <Prelude_Zzzzz> any difference in quality ?
[18:35:45 CET] <BtbN> Ask Nvidia
[18:35:53 CET] <BtbN> The onyl difference i know is the GUID passed to the driver.
[19:19:29 CET] <Prelude_Zzzzz> ic... hey something else that i think is an ffmpeg issue.. i am encoding and its working fine right now... problem is every once in a while i see a quick flash on the screen as if some other frame got inserted into the stream that doesn't belong there
[19:19:46 CET] <Prelude_Zzzzz> anyone know possibly why? its weird.. like a flash from another channel or something
[19:54:46 CET] <yumbox> hi, how can i get exact bitrate from aduio file?
[19:57:56 CET] <yumbox> when i do ffprobe it gives inaccurate: http://a.uguu.se/rqxqjn.png
[20:04:30 CET] <grublet> yumbox: you might want to try cahnging the analyze duration or probesize?
[20:04:42 CET] <grublet> afaik ffmpeg only takes a sample to determine the bitrate
[20:04:53 CET] <yumbox> can i make it accurate?
[20:05:10 CET] <yumbox> its no problem if the whole file needs to be read
[20:05:15 CET] <grublet> it depends on what you mean by accurate
[20:05:31 CET] <grublet> yumbox: try messing with the analyze duration and probe size switches
[20:05:59 CET] <grublet> yumbox: http://www.ffmpeg.org/ffprobe-all.html
[20:06:24 CET] <grublet> analyzeduration integer (input)
[20:06:24 CET] <grublet>     Specify how many microseconds are analyzed to probe the input. A higher value will enable detecting more accurate information, but will increase latency. It defaults to 5,000,000 microseconds = 5 seconds.
[20:06:32 CET] <grublet> probesize integer (input)
[20:06:32 CET] <grublet>     Set probing size in bytes, i.e. the size of the data to analyze to get stream information. A higher value will enable detecting more information in case it is dispersed into the stream, but will increase latency. Must be an integer not lesser than 32. It is 5000000 by default.
[20:06:41 CET] <yumbox> thanks!
[20:06:50 CET] <grublet> hopefully that works for you
[20:07:01 CET] <grublet> if not someone else might be better help
[20:10:32 CET] <yumbox> hmm, still gives same bitrate
[20:12:40 CET] <yumbox> can I just take the size of the .aac file and divide by its duration to get the average bitrate?
[20:12:49 CET] <yumbox> or do i need to get the header overhead into account?
[20:20:35 CET] <xkriegor> is this the right place to ask about avfoundation issues?
[20:21:29 CET] <xkriegor> well, ffmpeg avfoundation issues ofc
[20:26:22 CET] <grublet> yumbox: the bitrate may also include the container overhead
[20:26:41 CET] <yumbox> hmm :(
[20:26:48 CET] <grublet> so for example on an MP3 it may tell me 321Kbps or something similar
[20:27:51 CET] <grublet> or if it was encoded with VBR then chances are that average bitrate is due to fluctuations, even if it was 'set' for 160 for example
[20:28:27 CET] <grublet> yumbox: do you know know how the file was encoded? if it was cbr or 1 or 2 pass vbr?
[20:29:50 CET] <yumbox> i only know "AAC LC profile" and duration
[20:31:58 CET] <yumbox> i think i got it
[20:33:01 CET] <grublet> yumbox: how did you determine the solution?
[20:33:41 CET] <yumbox> mux into mp4
[20:33:44 CET] <yumbox> (m4a)
[20:34:00 CET] <yumbox> ffmpeg -i out.aac -c:a copy -bsf:a aac_adtstoasc out.m4a
[20:35:52 CET] <grublet> yumbox: i was going to suggest that
[20:36:09 CET] <grublet> im glad you figured it out ;)
[20:36:49 CET] <yumbox> but the size still doesnt make sense :(
[20:36:59 CET] <yumbox> im probably calculating wrong
[20:38:01 CET] <grublet> i'm out of ideas, so just stick around and someone else might know
[21:31:59 CET] <yumbox> redid size calculations and it seems like muxing to .m4a does give accurate bitrate
[21:34:01 CET] <yumbox> but not with mediainfo, only with ffprobe
[23:46:33 CET] <pmow> Hello!  I'm trying to convert MKVs to MP4 for device support.  Without a codec option, ffmpeg is happy to convert.  With a codec option, whether it's `-c:s copy` or `-c:s ass/ssa/subrip` I get the error Could not write header, incorrect codec params.  File: http://pastebin.com/qaJHpqLy
[23:56:16 CET] <pmow> sure
[00:00:00 CET] --- Thu Oct 29 2015


More information about the Ffmpeg-devel-irc mailing list