[Ffmpeg-devel-irc] ffmpeg.log.20170419

burek burek021 at gmail.com
Thu Apr 20 03:05:01 EEST 2017


[00:53:57 CEST] <kerio> is there a reference encoder for H.264
[00:56:11 CEST] <furq> http://iphome.hhi.de/suehring/tml/
[01:08:39 CEST] <jkqxz> The reference encoder for H.264 is a test platform, not something you would actually want to use.  It targets PSNR, and does it essentially as well as possible with full search and RDO across all usable features.  Unfortunately, that makes it (a) often visually inferior to the highest-quality settings of x264, because eyes don't see PSNR, and (b) laughably slow.
[01:15:25 CEST] <sshack> Is there an ffmpeg command which will stretch/shrink a video to a certain time?
[01:16:06 CEST] <sshack> I've got a bunch of clips, and I'd like them all to be exactly a certain length - I don't mind if they look wonky, this is for computer vision
[01:16:52 CEST] <llogan> see setpts filter
[01:18:31 CEST] <sshack> Okay, so if I only know the end desired length what am I doing then?
[01:20:24 CEST] <furq> -vf setpts=(x/y)*PTS
[01:20:33 CEST] <furq> where x is the desired length and y is the current length
[01:21:00 CEST] <sshack> Is there a stand in variable for the current length?
[01:21:12 CEST] <sshack> Or do I have to do some shell magic here?
[01:21:17 CEST] <furq> get it with ffprobe
[01:21:29 CEST] <sshack> Shell magic.
[01:22:00 CEST] <furq> http://vpaste.net/W10Ur
[01:22:40 CEST] <furq> x and y both have to be in seconds anyway
[01:23:15 CEST] <furq> it sure would be nice if the ffprobe invocation to get a single property was a bit shorter
[01:23:23 CEST] <sshack> Okay, I would *not* have figured that out.
[01:23:27 CEST] <furq> as opposed to being the longest command in the world
[01:24:36 CEST] <llogan> you can lengthen it by using -loglevel instead of -v
[01:24:36 CEST] <furq> stream_tags and stream_disposition being on by default is a particular annoyance
[01:25:03 CEST] <furq> those shouldn't show up by default if you're using -show_entries
[01:25:17 CEST] <furq> especially since stream_disposition isn't even documented
[01:26:37 CEST] <llogan> sshack: you'll also need "-select_streams v" so it doesn't return the audio duration as well
[01:26:43 CEST] <sshack> Does something like this look correct? http://vpaste.net/22K9Y
[01:26:46 CEST] <furq> oh yeah
[01:26:53 CEST] <llogan> or get the "format" duration
[01:27:04 CEST] <furq> sshack: you need to wrap the ffprobe command in $()
[01:27:25 CEST] <furq> also desiredlength needs to be a number, not a duration spec
[01:27:30 CEST] <sshack> furq Yeah I have to fixup my shell code. But the overall structure looks correct?
[01:27:44 CEST] <furq> and yeah also do the thing llogan said
[01:27:48 CEST] <llogan> jpg outputs look like an odd choice
[01:27:50 CEST] <sshack> so like desiredlength=2? or what?
[01:27:52 CEST] <furq> yeah
[01:28:14 CEST] <sshack> llogan Doing computer vision here. I need normalized video for short clips
[01:28:22 CEST] <furq> also probably quote $videoname
[01:29:18 CEST] <furq> also yeah that won't actually do anything if you're outputting to images
[01:29:31 CEST] <furq> you'll just get sequentially numbered images either way
[01:30:03 CEST] <sshack> http://vpaste.net/EkLMw
[01:30:11 CEST] <sshack> Won't actually do anything?
[01:30:24 CEST] <furq> jpegs don't have timestamps, for obvious reasons
[01:30:32 CEST] <sshack> Well I want to guarantee I get <somenumber of frames>
[01:30:40 CEST] <furq> oh
[01:30:45 CEST] <furq> in that case you want to use -r to drop frames
[01:31:06 CEST] <sshack> will -r work with varying input clips?
[01:31:07 CEST] <llogan> you have an errant quote
[01:31:20 CEST] <llogan> run it through shellcheck.net
[01:31:29 CEST] <furq> it will but you'll need to get the framerate with ffprobe as well afaik
[01:32:06 CEST] <furq> so run ffprobe again but replace duration with avg_frame_rate
[01:32:20 CEST] <sshack> aahrg! Thank you guys for helping out. There's no way I would have figured this out.
[01:32:31 CEST] <llogan> and get rid of -show_streamsd
[01:32:38 CEST] <llogan> *-show_streams
[01:33:39 CEST] <sshack> http://vpaste.net/IaAuV
[01:33:57 CEST] <furq> do you actually need an even spread of frames from throughout the video
[01:34:14 CEST] <furq> if you can just clip the first n frames then you don't need to do any of this
[01:34:59 CEST] <sshack> furq  I would prefer an even spread. Extremely annoying, but it doesn't work otherwise.
[01:35:14 CEST] <furq> well yeah
[01:36:24 CEST] <llogan> get rid of -show_streams again. then you can probably omit ":stream_tags=0:stream_disposition=0" as well.
[01:36:35 CEST] <furq> replace -vf setpts with -vf "fps=$fps*($desiredlength/$currlength)"
[01:36:48 CEST] <furq> or $framerate or whatever you're calling it
[01:37:13 CEST] <furq> that might end up being off by one though
[01:37:25 CEST] <llogan> and you might as well replace -s with the scale filter
[01:37:27 CEST] <furq> it seems like there should be a better way to do it
[01:37:34 CEST] <furq> i can't think of one though
[01:37:58 CEST] <sshack> http://vpaste.net/W3quA
[01:38:29 CEST] <sshack> -s works, so lets leave that for now. No point in poking a hornets nest.
[01:39:31 CEST] <llogan> it's easy: "fps=$fps*($desiredlength/$currlength),scale=320:-1"
[01:40:08 CEST] <sshack> Oh. Okay, I can take that risk
[01:43:17 CEST] <sshack> Does this look kosher?
[01:43:18 CEST] <sshack> http://vpaste.net/6R52K
[01:46:40 CEST] <sshack> "Argument 'Untitled00000133.mov' provided as input filename, but 'stream=duration' was already specified."
[01:47:18 CEST] <sshack> That's from trying to get the current length
[01:48:03 CEST] <sshack> Oh typo on that line.
[01:48:18 CEST] <sshack> Now I get Argument 'Untitled00000133.mov' provided as input filename, but 'avg_frame_rate' was already specified from the frame rate gathering part
[01:48:35 CEST] <llogan> another typo
[01:49:41 CEST] <llogan> stream= avg
[01:50:22 CEST] <sshack> Yeah got that.
[01:50:27 CEST] <sshack> Now I'm up against shell quoting.
[01:50:29 CEST] <sshack> http://vpaste.net/U6dcb
[01:50:53 CEST] <llogan> use shellcheck
[01:54:33 CEST] <sshack> ahh fps.
[01:54:47 CEST] <sshack> so If I'm normalizing the length of a video, what is FPS doing?
[01:58:02 CEST] <sshack> I think I have it running btw, I'm just asking for curiosity at this point.
[02:00:09 CEST] <sshack> Ahh. These clips are coming up short.
[02:00:21 CEST] <sshack> I've got a bunch of < 40 frame clips here.
[02:05:40 CEST] <sshack> Oh no my mistake. Needed to update the driver script.
[02:27:10 CEST] <sshack> If I want just a fixed 20fps ffmpeg -r 20 would do that or not?
[02:28:16 CEST] <sshack> or is that -r 0.05
[03:27:23 CEST] <sshack> Got it working. Thank's very much for the help everyone!
[07:29:12 CEST] <ariay> I want to stream input to ffmpeg -i, but that never appears to work. The input stream always has to close before ffmpeg actually does anything. Does anyone have experience with streaming input to ffmpeg?
[07:29:32 CEST] <ariay> If I use pipe:0, I must close the pipe, if I use tcp://ip:port, I must close the connection.
[07:30:47 CEST] <ariay> I am feeding raw pcm_s16le frames to ffmpeg and want to do transformations on frames instead of the entire audio file, is this even possible?
[10:03:41 CEST] <masterdd> Hi everybody, I was searching a way to dump AVI stream-header data chunk (FOURCC code 'strd') using ffmpeg, does anyone kwown how to do it?
[10:20:13 CEST] <marbar> hey all. i have a 14-second video recorded with Chrome's MediaRecorder (file: http://mkult.co/wmtest/src.webm, ffprobe: http://mkult.co/wmtest/ffprobe.txt). as you can see, it doesn't play correctly.
[10:20:24 CEST] <marbar> i tried "ffmpeg -i src.webm out.mp4" (resulting in file: http://mkult.co/wmtest/out.mp4 with output: http://mkult.co/wmtest/output.txt). both files are identically unplayable.
[10:20:33 CEST] <marbar> i think this is at least in part because MediaRecorder-generated webm files don't have cues -- is this right? do you any of you fine folks know how to insert cues, or whether this is the right approach? thanks, any help is greatly appreciated :)
[10:42:54 CEST] <dystopia_> i think "ffmpeg -i src.webm out.mp4" is basically just a copy operation to the diffrent container
[10:43:00 CEST] <dystopia_> try encode it
[10:45:06 CEST] <dystopia_> ffmpeg src.webm -c:v libx264 -preset slow -acodec copy out.mp4
[10:46:36 CEST] <marbar> dystopia_, thanks dude! i'll give it a shot.
[10:48:46 CEST] <marbar> dystopia_, that makes it complain "Could not find tag for codec opus in stream #1..."
[10:48:57 CEST] <dongs> opus is audio
[10:49:03 CEST] <marbar> dystopia_, this is because (i think) the audio/video streams are weirdly switched
[10:49:04 CEST] <dongs> the line he pasted only does video.
[10:49:06 CEST] <marbar> yeah exactly
[10:49:14 CEST] <dongs> so you want to convert the audio too
[10:49:15 CEST] <marbar> this webm file has the audio stream before the video stream
[10:49:35 CEST] <marbar> oh i'll try adding something for that, good call dongs
[10:50:07 CEST] <dongs> no, order of them doesnt matter
[10:50:13 CEST] <dongs> just add a thing to transcode audio to aac or something too
[10:50:22 CEST] <dongs> -acodec aac or someshit< i forgot which one is hte better sounding
[10:50:33 CEST] <dongs> nothing plays opus since it s opensores
[10:50:49 CEST] <marbar> haha
[10:50:55 CEST] <marbar> dongs i added -c:a aac
[10:51:00 CEST] <marbar> and it encoded, throwing an error
[10:51:09 CEST] <dongs> probably need some freetard version of aac
[10:51:12 CEST] <marbar> and the resulting file was the same as the others
[10:51:27 CEST] <dongs> i only know -acodec but i think tis same as -c:a
[10:51:49 CEST] <marbar> dongs, do you know  how to inject cues into a webm file?
[10:52:00 CEST] <marbar> or fix messed up durations? i think it's something to do with that...
[10:52:13 CEST] <marbar> bc this is just a file fresh out of MediaRecorder (html5)
[10:52:43 CEST] <marbar> and i've read some stuff indicating these files need to be fixed in some way (related to cues and durations)
[10:53:21 CEST] <dongs> sorry i only dealt with duration stuff with mpegts type stuff, no idea how its done with webm/mp4. but I think defaults should be sane, when you go vp9->h264 and opus->aac and remux to mp4
[10:53:46 CEST] <marbar> used this command: ffmpeg -i src.webm -c:v libx264 -preset slow -c:a aac enced.mp4
[10:54:21 CEST] <marbar> but any player i use (mplayer, ffplay, etc) results in a file that's mostly frozen on one frame, with some playing before and after. weird :'(
[10:55:02 CEST] <marbar> yeah, and they just keep playing, as well -- they never end. infinite duration is something i've read about associated w/ MediaRecorder-recorded webm's.
[10:55:14 CEST] <marbar> (frozen on one frame, that is)
[10:55:46 CEST] <dongs> .webm is the failing-to-encode source right?
[10:55:49 CEST] <dongs> k grabbing it and looking
[10:56:11 CEST] <marbar> thanks :) you're my hero dongs
[10:57:14 CEST] <dongs> marbar: the problem is your original webm is acutalyl damaged/trashed
[10:57:15 CEST] <marbar> (and the log link i posted above, output.txt, is of the first ffmpeg run, w/o the encoding options, but the result is the same)
[10:57:21 CEST] <marbar> oh gross
[10:58:11 CEST] <dongs> there's either chunks of data missing (frame/write drops?) or your encoder is writing garbage data
[10:58:27 CEST] <marbar> well, every blob i've gotten from MediaRecorder seems to be messed up in this way. it's recorded in segments and stitched together in a single blob in js, then downloaded. i'll try a simpler test.
[10:58:38 CEST] <dongs> the missing data cuases timestamp jumps which is why it sorta plays for a sec then freezes then palys some more
[10:58:41 CEST] <marbar> it's the chrome mediarecorder
[10:58:53 CEST] <marbar> oh gross. is there any way to smooth this sort of thing out?
[10:58:54 CEST] <dongs> i think youre nto combining chunks correctly
[10:59:03 CEST] <dongs> whatever the "correct" way should be
[10:59:05 CEST] <marbar> ok, i'll research that
[10:59:07 CEST] <marbar> thanks dude :)
[10:59:34 CEST] <dongs> i wonder if you stitched them out of order
[11:01:06 CEST] <marbar> maybe. i'm using a function from called ConcatenateBlobs in the MediaStreamRecorder library. it's actually the only reason i haven't dropped the library yet, as most of the shimmed functionality is up to speed in browsers at this point.
[11:01:10 CEST] <marbar> it's pretty dumb of me
[11:02:09 CEST] <dongs> marbar: is chunk size 512k by any cahnce
[11:02:44 CEST] <marbar> i don't think so. i think it's smaller.
[11:02:46 CEST] <marbar> how come?
[11:03:08 CEST] <dongs> nah just wondering. right, it cant be 512k cuz the damage is much earlier
[11:03:15 CEST] <dongs> one of the first frames is already cut resulting in decode fail
[11:03:23 CEST] <dongs> cut = incomplete
[11:03:27 CEST] <marbar> oh gross
[11:03:30 CEST] <marbar> that's pretty wacky
[11:03:36 CEST] <marbar> yeah i'm gonna drop this questionable lib
[11:09:58 CEST] <marbar> holy crap dongs
[11:10:01 CEST] <marbar> you're so right
[11:10:20 CEST] <marbar> instead of calling that whatever it was function
[11:10:33 CEST] <marbar> i just said "new Blob(the_blob_array)"
[11:10:35 CEST] <marbar> and it was fine
[11:10:40 CEST] <marbar> thanks dude, thanks a million :)
[11:11:54 CEST] <dongs> ^_^
[12:18:01 CEST] <momomo> how efficient is ffmpeg compared to other tools for encoding?
[12:19:59 CEST] <dongs> over 6.66% more efficient
[12:20:17 CEST] <dongs> up yo 66.6% if you enable extra otpions
[12:23:35 CEST] <dystopia_> lots of other tools actually use ffmpeg
[12:23:51 CEST] <dystopia_> they just have a fancy gui and try to sell them selves to you
[12:30:49 CEST] <momomo> i once came across a comparison but unable to find it
[12:30:55 CEST] <momomo> spent like 20 minutes looking for it now
[12:54:43 CEST] <momomo> what are cheap cloud live encoding services? I am thinking thousands of streams 24h per day
[13:00:15 CEST] <BtbN> good luck finding that for cheap
[13:00:47 CEST] <BtbN> And ffmpeg just uses the standard-encoders for most codecs, so comparing ffmpeg to something else doesn't really make sense
[13:02:15 CEST] <dongs> sure it does
[13:02:25 CEST] <dongs> you can take reference h264 encoder from mpeg-la or wahtever
[13:02:29 CEST] <dongs> vs h264 stuff in ffmpeg
[13:02:34 CEST] <dongs> one will be encoding at liek 1fps
[13:02:37 CEST] <dongs> while the other at over 9000
[13:02:44 CEST] <dongs> both outputting same h264
[13:02:45 CEST] <BtbN> There is no h264 stuff in ffmpeg
[13:02:56 CEST] <dongs> that was not my point
[13:03:00 CEST] <BtbN> it just uses x264 or some other encoder, mostly hardware encoders besides x264
[13:03:28 CEST] <dongs> reference encoder <produced by patent/etc group A> vs <encoder outputting same bitstream in ffmpeg>
[13:03:47 CEST] <BtbN> you won't find two h264 encoders that produce the same bitstream for the same input though
[13:03:52 CEST] <dongs> none of hardware encoders to crf
[13:04:02 CEST] <dongs> to->do
[13:04:17 CEST] <BtbN> crf is a x264 implementation specific thing
[13:04:23 CEST] <BtbN> the hardware decoders to cqp
[13:04:25 CEST] <BtbN> *encoders
[13:04:48 CEST] <momomo> dongs: not sure what side you are on? ffmpeg good?
[13:05:07 CEST] <dongs> i'm just trolling so whichever side is easier to troll I'm there.
[13:06:15 CEST] <BtbN> ffmpeg is neither good nor bad, it's just a unified interface for a lot of encoders, and in some cases like for example aac it's also an encoder itself
[13:06:27 CEST] <BtbN> On the decoding side there is a lot more stuff implemented directly in ffmpeg though
[13:07:42 CEST] <momomo> fcuk .. i cant find the alternative i once ran into ... something with outputstream .. a commercial software that claimed to be much more performant
[13:38:32 CEST] <momomo> finally found the comparison
[13:38:33 CEST] <momomo> https://software.intel.com/en-us/articles/video-encoding-using-the-integrated-intel-hd-graphics
[13:38:56 CEST] <momomo> look at the time for ffmpeg
[13:39:01 CEST] <momomo> look at the time for streambuilder
[13:39:30 CEST] <BtbN> that compares the Intel Hardware-Encoder... with itself
[13:40:01 CEST] <BtbN> Or rather the various implementations of it
[13:40:55 CEST] <BtbN> It's from 2014, I don't think ffmpeg even supported qsv/libva back then, so they are probably comparing their hardware encoder against x264, which is unsurprisingly slower
[13:42:06 CEST] <momomo> it says "StreamBuilder is based on libavcodec" ...
[13:42:19 CEST] <momomo> does ffmpeg base on that for 264 ?
[13:47:20 CEST] <BtbN> libavcodec is an ffmpeg library.
[14:03:14 CEST] <IntruderSRB> hey guys - anyone aware of some MP4 inspector application for MacOS?
[14:03:33 CEST] <IntruderSRB> other than mp4dump/mp4info from Mp4Box
[14:09:58 CEST] <JEEB> IntruderSRB: boxdumper from L-SMASH
[14:11:18 CEST] <IntruderSRB> hey @JEEB thanks
[14:11:26 CEST] <IntruderSRB> looks like it's CLI only for Mac or am I missing something?
[14:19:38 CEST] <ritsuka> if you want a gui there is atom inspector
[14:28:00 CEST] <JEEB> IntruderSRB: it's cli-only. `boxdumper --box FILE | less` is how I usually use it
[14:28:10 CEST] <JEEB> because less lets me scroll up/down within the buffer nicely
[14:28:27 CEST] <IntruderSRB> well for cli dumps I was using mp4dump from Mp4Box
[14:28:30 CEST] <IntruderSRB> works pretty nice
[14:28:55 CEST] <IntruderSRB> but I need to compare two mp4 fragments for the same data to find differences
[14:28:59 CEST] <IntruderSRB> which are a bit more subtle
[14:39:25 CEST] <zanoni> hi, I'm trying to create a rtpm strean using ffmpeg, using -i pipe:0 as input and feeding image data to it, but I'm getting a very slow frame rate and the streaming don't start, can someone help me? this is the command and the output, it's a qt project and I'm starting ffmpeg using QProcess and writing the images to the process input
[14:39:52 CEST] <zanoni> *rtmp
[14:40:22 CEST] <zanoni> forgot the paste: https://pastebin.com/fUNNhRaT
[15:02:42 CEST] <JEEB> IntruderSRB: then dump the things into files and `diff -u file1 file2`
[15:03:42 CEST] <JEEB> (possibly using less too so you have a scrollable buffer because a lot of the offsets will be different)
[15:36:01 CEST] <IntruderSRB> @JEEB yep ... that's what I was doing lately :) btw. the streams I'm comparing is Progressive vs. Interlaced H264
[15:36:31 CEST] <IntruderSRB> I have unique pipeline/muxer for both that adds DRM on top of the stream and encrypts it
[15:36:40 CEST] <IntruderSRB> when I pass progressive video it works perfectly
[15:37:02 CEST] <IntruderSRB> for interlaced I get MEDIA_DECODE_ERR and the only difference in input streams that I see
[15:37:13 CEST] <IntruderSRB> is the presence of SEI nal units in the interlaced stream
[15:37:18 CEST] <IntruderSRB> which I leave unencrypted
[15:37:35 CEST] <IntruderSRB> anyone feeling like sharing some thoughts on the differences?
[17:46:04 CEST] <doverZ> hi, having a issue when playing a video on playstation3... on pc is ok but on ps3 the audio plays before than it should. Searched on internet little and found a solution using ffmpeg, applied but its still not good. The command ive used is this https://pastebin.com/VjAgP0Yb  ... tried to set the delay as that command shows and also in timer format HH:MM:SS.xxx with the same results. When i check the mediainfo of the original mkv
[17:46:04 CEST] <doverZ> file it has 5 seconds and 100 ms of delay for the audio, but if i check the mediainfo of the converted mp4 result files with or without delay, it doesnt show up the delay data...its like if its doing nothing
[18:15:48 CEST] <pihpah> Why ffprobe does not show audio bitrate?   Stream #0:1(eng): Audio: aac (HE-AAC), 48000 Hz, stereo, fltp
[18:15:52 CEST] <pihpah> In some cases
[18:24:01 CEST] <thebombzen> what version
[18:24:16 CEST] <thebombzen> also the bitrate is one line up often
[18:24:32 CEST] <pihpah> ffmpeg version 3.2.2
[18:26:02 CEST] <furq> it depends on the container
[18:26:32 CEST] <pihpah> furq:  https://pastebin.com/cwShYaVk
[18:26:39 CEST] <thebombzen> yea. some containers like matroska dont' have that information stored
[18:26:57 CEST] <thebombzen> other ones like mp4 do
[18:27:23 CEST] <pihpah> thebombzen: you are right, that's an mkv container.
[18:28:59 CEST] <thebombzen> pihpah: try this: ffmpeg -i input.mkv -map a -c copy -f adts - 2>/dev/null | ffmpeg -f aac -i -
[18:29:01 CEST] <pihpah> So what would happen if I specify a higher audio bitrate when transcoding? Would ffmpeg just ignore that value?
[18:29:16 CEST] <thebombzen> when transcoding the input bitrate is unimportant because it's decoded anyway
[18:29:35 CEST] <thebombzen> you have to specify the bitrate when transcoding every time, and if you don't, it'll default to 128 kbps
[18:31:01 CEST] <pihpah> Duration: N/A, bitrate: 71 kb/s the output of your command
[18:31:34 CEST] <pihpah> Isn't it to low?
[18:31:43 CEST] <pihpah> too*
[18:32:07 CEST] <thebombzen> no
[18:32:19 CEST] <thebombzen> that's HE-AAC, 71 kb/s sounds about right
[18:33:01 CEST] <thebombzen> it's not high quality audio you have there. it's a fairly low bitrate. but it's not so low that it looks "wrong"
[18:33:38 CEST] <thebombzen> I mean, you wanted to know the bitrate so I gave you a command to do it
[18:34:51 CEST] <pihpah> ok, I suppose I have just to copy that audio stream leaving it untouched. Usually I convert audio streams to 64, 128 and 192 bitrates for different video resolutions.
[18:36:26 CEST] <pihpah> thank you
[19:06:09 CEST] <Mandevil> My encodes with libx264 always freeze after few minutes... why? https://pastebin.com/H3vUZx0p
[19:06:25 CEST] <Mandevil> Encodes in libx265 or prores are fine.
[19:07:22 CEST] <thebombzen> also you realize you have -pass 1, right?
[19:07:29 CEST] <Mandevil> Yes.
[19:09:08 CEST] <Mandevil> thebombzen: That doesn't mean it should freeze mid-encode.
[19:09:23 CEST] <thebombzen> you're right, it shouldn't, but perhaps you should past the complete console output
[19:09:42 CEST] <thebombzen> otherwise we'll be like "well, it works when we run that command"
[19:09:54 CEST] <Mandevil> Can I log the ouput into file while also keeping the display output?
[19:10:09 CEST] <Mandevil> Like tee in Unix, but in Windows.
[19:10:17 CEST] <thebombzen> http://google.com/
[19:25:54 CEST] <Mandevil> OK, here's the complete log: https://pastebin.com/KugL3Wqs
[19:26:09 CEST] <Mandevil> It ends with me ctrl-c'ing the encode because it just gets stuck.
[20:11:44 CEST] <thebombzen> that doesn't look stuck
[20:12:07 CEST] <Mandevil> thebombzen: CPU load drops and it does nothing.
[20:12:38 CEST] <thebombzen> what's in mylist.txt
[20:13:06 CEST] <Mandevil> https://pastebin.com/45KerYsC
[20:14:27 CEST] <thebombzen> what's in each of those files
[20:14:36 CEST] <thebombzen> also did you know AVI is cat-compatible
[20:14:43 CEST] <thebombzen> can just literally concatenate them together
[20:15:49 CEST] <Mandevil> thebombzen: The AVIs are MJPEGs from a dash cam.
[20:16:00 CEST] <Mandevil> thebombzen: Again, libx265 had no problem encoding the same exact thing.
[20:16:18 CEST] <thebombzen> then it sounds like your libx264 build is bad
[20:16:24 CEST] <thebombzen> and you should get a recent version
[20:18:23 CEST] <Mandevil> OK, downloaded the latest and trying again.
[20:18:26 CEST] <furq> even if one of the inputs was corrupt, ffmpeg shouldn't just freeze
[20:19:02 CEST] <furq> but yeah if it works with other encoders then libx264 is probably to blame
[20:19:13 CEST] <thebombzen> I missed the "works with other encoders" part
[20:26:37 CEST] <Mandevil> Hm, 15 minut and going...
[20:26:39 CEST] <Mandevil> minutes
[20:31:21 CEST] <Mandevil> 24 minutes... this looks much better.
[20:34:47 CEST] <furq> zeranoe builds with x264 from git head, so maybe you just got a bad commit
[20:35:26 CEST] <furq> i'm not sure if the builds are tested for issues like that
[20:36:47 CEST] <JEEB> x264 having issues is something very, very rare
[20:36:59 CEST] <JEEB> of course worth checking if it really is x264, but still
[20:58:43 CEST] <Mandevil> First pass completed fine.
[21:38:20 CEST] <petecouture> Question for you video designers out there. Say I have two videos I want to composit together. In AfterEffects I can do something like an Iris transition from video A to video B where both videos overlay for a 1/2 second as the transition takes place. Is something like that possible in ffmpeg?
[21:42:56 CEST] <petecouture> To clarify I'm not looking for ffmpeg do animate the iris transition
[21:43:42 CEST] <petecouture> I'm assuming I will render out the iris transition in aftereffects and maintain an alpha channel in the video of the iris
[21:47:29 CEST] <furq> petecouture: overlay will do that
[21:51:21 CEST] <petecouture> @furq awesome mate thank you as always for your help!
[22:06:56 CEST] <ChocolateArmpits> petecouture, you may want to look into vapoursynth or avisynth in the future. They are essentially scriptable compositors
[22:25:20 CEST] <petecouture> @ChocolateArmpits thank you, I will look into them. My client has a lot of crazy video transitions he's looking to get scripted out so the source media can be dynamically changed. Essentially this is going to be a hybrid of AERender doing all the heavy graphic rendering and ffmpeg doing the compositing and encoding. I was just wondering if two videos ( one with an alpha channel ) can be composited.
[22:50:19 CEST] <sshack> Hi guys. Thanks for helping me yesterday. I have a question about working with videos in python.
[22:50:53 CEST] <sshack> Are there any python libraries to drive ffmpeg (or anything else)? I'd like to extract frames and rescale them to do computer vision work.
[22:52:16 CEST] <JEEB> any pre-made mappings are most likely outdated by now, so I'd say mappings generated from your current FFmpeg headers/libs a la VLC's script is probably the least bad way
[22:52:21 CEST] <JEEB> or you take a look into vapoursynth
[22:52:42 CEST] <JEEB> which is a frameserver the main controller of which is python3
[22:54:32 CEST] <sshack> python3 dang. I'm doing TF code in 2.7
[22:55:22 CEST] <thebombzen> you should be using python3 anyway
[22:55:28 CEST] <thebombzen> this is just a good opportunity to learn it
[22:57:16 CEST] <sshack> I know, I just tend to be shackled by older libraries.
[23:01:37 CEST] <Aerroon> hey guys, do haswell CPUs (4790k, 4670k) support quicksync's LA_ICQ properly?
[23:02:27 CEST] <furq> sshack: there's pyav but idk how good it is
[23:02:33 CEST] <furq> it looks to be one of the better bindings, but that's not saying much
[23:03:03 CEST] <JEEB> if something came out of rust, the bindgen thing is nice
[23:03:15 CEST] <JEEB> I'm really liking the idea of generating bindings as part of the build process
[23:03:53 CEST] <sshack> furq Uhm, if it's that simple to use, and has rescale it's *exactly* what I want https://pastebin.com/W0f6By6K
[23:04:08 CEST] <furq> that sure does look simple
[23:04:40 CEST] <furq> assuming you can install it in the first place
[23:04:41 CEST] <JEEB> almost looks like ffms2 level of stuff
[23:04:58 CEST] <JEEB> (which is another nice library on top of FFmpeg's which focuses on frame-exactness and other things)
[23:04:59 CEST] <sshack> It even output's PIL images. *swoon*
[23:05:21 CEST] <furq> i considered using it but i was put off by it admitting in the readme that it was a pain to install
[23:05:34 CEST] <furq> it's probably fine if you don't need this to be distributable, but i did
[23:08:11 CEST] <sshack> Distribution isn't a problem for me.
[23:08:21 CEST] <sshack> furq Thank you very much. I think this might be gold for me.
[23:17:29 CEST] <Aerroon> okay nevermind, question still stands
[23:17:47 CEST] <Aerroon> do the haswell CPUs work fine with LA_ICQ with QSV?
[23:18:40 CEST] <BtbN> QSV in general doesn't work fine. What even is LA_ICQ?
[23:19:17 CEST] <Aerroon> look ahead intelligent constant quality
[23:19:56 CEST] <Aerroon> well, i have the question more due to OBS, ICQ works fine on my 4790k, but LA_ICQ does not seem to actually take any options into account
[23:20:08 CEST] <Aerroon> and the same thing happened to a few other people i saw and asked questions from
[23:20:29 CEST] <Aerroon> so i was wondering if anybody knew if maybe the igpus themselves don't support it properly or if it's an obs issue
[23:27:38 CEST] <sshack> furq I can see what you mean about "hard to install" now. Tracking down -dev libraries in ubuntu. Brings back not-gear memories.
[00:00:00 CEST] --- Thu Apr 20 2017


More information about the Ffmpeg-devel-irc mailing list