[Ffmpeg-devel-irc] ffmpeg.log.20181125

[burek]

[02:30:14 CET] <markweston> what is the smallest OS (in total lines of code) that has ffplay ported to and can play x254 video @ 720p?
[06:46:23 CET] <Woodpecker> I need to take a directory full of about 3000 audio files, with wildly different volumes with many peaks and valleys, and adjust the audio so it is all within a very uniform range. That means, no peaks or valleys are allowed to exist outside the range. I dont know what the term for this process is, or where to start.
[07:26:40 CET] <sinclair> Could not write header for output file #0 (incorrect codec parameters ?): Invalid argument
[07:28:59 CET] <sinclair> sorry Woodpecker
[07:29:24 CET] <Woodpecker> sinclair: I think I found it, loudnorm
[07:30:01 CET] <sinclair> Woodpecker: I have a bit of a issue issue with ffmpeg myself, im trying to pipe images in over stdin, and pipe h264 (mp4) video out over stdout.
[07:30:31 CET] <Woodpecker> sinclair: I actually have a bit of experience with that.
[07:30:36 CET] <Woodpecker> sinclair: are you doing a slideshow?
[07:31:08 CET] <sinclair> Woodpecker: well, i have a bunch of rendered (ray traced frames), these are computed at 24fps, and i have thousands of them.
[07:31:20 CET] <sinclair> Woodpecker: am using image2pipe, but im not really sure what i should be doing
[07:31:32 CET] <sinclair> Woodpecker: there is no 'audio' in my case.
[07:32:54 CET] <Woodpecker> sinclair: I dont use quite so many frames. My videos might have 100-200 frames for an hour long video. For concatenating them... mmm let me check what I was using.
[07:32:55 CET] <sinclair> Woodpecker: the idea is to be able to feed these frames into ffmpeg, have it encode them as mp4, then stream them back out over stdout (which get persisted on a web server), the whole thing is distributed, and i can't rely on sourcing images from disk using patterns, its all got to be done over stdio.
[07:33:13 CET] <sinclair> Woodpecker: any examples would be great
[07:35:56 CET] <Woodpecker> sinclair: unfortunately my code has been lost in a tragic wipe, but if I recall, I was using imagemagick to create the images, and for the timing, I used an ffmpeg script file
[07:36:50 CET] <Woodpecker> and now I got to go
[07:36:52 CET] <Woodpecker> sorry
[07:37:13 CET] <sinclair> Woodpecker: alright, well, thanks anyway
[08:01:46 CET] <sinclair> is anyone able to help me with this command
[08:01:48 CET] <sinclair> 'ffmpeg -f image2pipe -vcodec png -framerate 24 -f mp4 -i - -r 24 -s 1024x768 -f mp4 -crf 15 -pix_fmt yuv420p -preset veryslow pipe:0'
[08:01:58 CET] <sinclair> i have no idea what im doing \o/
[08:18:36 CET] <sinclair> if im piping images in over stdin, how do i tell ffmpeg that its received the last image and should process and exit?
[08:18:51 CET] <sinclair> i think there are likely images still left in the codec buffer
[08:22:21 CET] <sinclair> is there a good source of information for using ffmpeg?
[08:25:34 CET] <sinclair> hello?
[13:13:03 CET] <ProN00b> is there a way to generate HLS/DASH segments on demand? (like if someone wants to skip to the last 10 seconds of a video, that only the segments around that get transcoded?)
[13:41:55 CET] <JEEB> ProN00b: make an API client and generate the segments on demand. for live the most usual design is to have an archive which is being fed by a live encoder from one side, and then you on-demand generate what clients request
[13:42:21 CET] <JEEB> for VOD if your content is already properly encoded then you already have your archive
[14:29:51 CET] <FurretUber> Hi, I'm trying to encode video using vp8_vaapi but the encoding is failing: http://termbin.com/75k7 The vainfo output can be seem at: http://termbin.com/8i1g
[14:31:00 CET] <FurretUber> The CPU is an i3-6100U and the GPU is an Intel HD Graphics 520. With hevc_vaapi a very similar error happens while with h264_vaapi it works
[14:40:28 CET] <jkqxz> FurretUber:  Is that dynamic linker path override making it pick a different driver, perhaps?
[14:40:54 CET] <jkqxz> The log with -v debug will show what driver it actually loaded.
[14:47:23 CET] <FurretUber> Looks like it found the driver in the right location: http://termbin.com/50tm I'm not replacing libraries related to VAAPI, only software encoders (libopus, libvpx, libx264)
[14:48:30 CET] <jkqxz> Ah.  There isn't a low-power encode entrypoint for VP8.  Remove the -low_power option.
[14:52:19 CET] <FurretUber> Now it works. Thank you. As it was showing in -h encoder=vp8_vaapi I thought it was working for it too
[14:55:09 CET] <jkqxz> It depends on the driver.  A driver could offer that, but from vainfo your version only does for H.264.  (Some Intel drivers also offer it for H.265, and maybe more in future.)
[14:56:04 CET] <FurretUber> From my tests now, only H264 and MPEG2 works with low-power option
[17:58:52 CET] <preyalone> anyone else notice ffmpeg -i running more slowly when overwriting an existing output file, compared to creating a new output file?
[20:02:36 CET] <pong> that sounds OS-specific
[23:28:18 CET] <OnkelTem> Oh hi
[23:28:28 CET] <OnkelTem> Glad to see this channel alive
[23:29:12 CET] <OnkelTem> I'm on Ubuntu 18.04 and trying to convert some MP2 audiotracks to AAC. And I'm curious how can I do it using libfdk-aac cuz it's available.
[23:29:18 CET] <OnkelTem> it's NOT available
[23:29:27 CET] <OnkelTem> So... what options guys do I have now?
[23:30:35 CET] <FurretUber> There is the aac encoder. Not sure about how good the aac encoder from the FFmpeg version Ubuntu ships is
[23:30:53 CET] <OnkelTem> you mean -a:c aac?
[23:30:59 CET] <FurretUber> Yes
[23:30:59 CET] <OnkelTem> -c:a*
[23:31:44 CET] <OnkelTem> Good, I'm using exactly that now. Just tried to follow steps from here https://superuser.com/a/370637/122672
[23:31:58 CET] <OnkelTem> For some reason libfdc_aac was mentioned first
[23:32:11 CET] <OnkelTem> and it said something about possible issues with source VBR
[23:32:47 CET] <OnkelTem> I haven't faced any though :)
[00:00:00 CET] --- Mon Nov 26 2018