[Ffmpeg-devel-irc] ffmpeg.log.20160509

[burek]

[00:33:33 CEST] <CoJaBo> Can anyone decipher "Past duration 0.679985 too large" for me? It seems to be causing dropped(==) frames every 2 seconds
[00:39:41 CEST] <jonascj> Hi all. I have an mp4 file for which "ffmpeg -i file.mp4" outputs "Stream #0:0(eng): Audio: aac (mp4a / 0x6134706D), 22050 Hz, stereo, fltp, 62 kb/s" among other things (like chapters). Is that 62kbit/s?
[00:40:04 CEST] <c_14> yes
[00:40:38 CEST] <jonascj> just like mp3 audio would typically be above 128kbit/s?
[00:40:56 CEST] <c_14> yes
[00:42:17 CEST] <jonascj> of course the codec here is acc, which should perform better than mp3 at the same bitrate, but still. This is audible.com's best quality, converted to mp4, not reencoding... kind of disappointing.
[00:42:25 CEST] <jonascj> I had to admit though, that it sounds okay
[00:43:32 CEST] <klaxa> it may be variable bitrate
[00:43:59 CEST] <c_14> If it's spoken voice, 62kb/s is relatively normal
[00:44:19 CEST] <klaxa> that too, would fit the 22050 Hz too
[00:44:26 CEST] <jonascj> it is spoken voice, an audiobook
[00:45:00 CEST] <jonascj> does halving the sample rate effectively double the bitrate?
[00:45:34 CEST] <jonascj> it feels like it should save space somewhere. It cuts the filesize in half I suppose
[00:48:40 CEST] <jonascj> well if the quality should remain the same, and the sampling frequency should double (to 44100Hz) the bitrate would have to double as well. So I answered my own question I think. I.e. if it was 44100Hz then it would probably have been 124kb/s which is more like the mp3 128kb/s @ 44100Hz I am use to
[00:50:36 CEST] <jonascj> Thanks for clarifying this.
[01:07:57 CEST] <klaxa> jonascj: i'm not so sure about halving it, since i don't know how audio codecs work
[01:08:47 CEST] <klaxa> i imagine they don't compress every sample for itself
[01:14:59 CEST] <iive> jonascj: mos/al mp3 encoders cut frequencies above 16kHz, so you are already limited to something like 32kHz sampling.
[08:35:20 CEST] <computer> hey people. can i convert from .SWF to .AVI or .M4P with high quality? when i use the -i command to convert to .AVI it is not very high quality
[08:36:10 CEST] <computer> i use the command ffmpeg -i file.swf file.avi
[08:41:27 CEST] <furq> replace avi with mp4
[08:42:05 CEST] <computer> furq: i get only aufio
[08:42:23 CEST] <computer> also when i convert to AVI i get a GREAT ZOOMED video of a mouse moving
[08:43:41 CEST] <computer> furq: i think ffmpeg is chaging the screen size
[08:49:39 CEST] <computer> furq: i am having an issue with converting the file correctly, i think because it id flash and it has a mouse pointer moving and ffmpeg is only converting the mouse movement. can you assist me? i have 300 files to convert i can send you one and see which commads can best help me?
[09:05:30 CEST] <computer> any help
[09:05:31 CEST] <computer> ?
[09:08:44 CEST] <computer> i am getting the following error:
[09:08:46 CEST] <computer> Input stream #0:0 frame changed from size:27x6 fmt:argb to size:87x38 fmt:argb
[09:32:05 CEST] <neuro_sys> doc/examples helped me understand the basics of io, demuxing, filtering and whatnot, and now I need to look into how to feed a raw in memory image into the filter. (Would that be by reading the raw data into an AVFrame struct?
[09:59:37 CEST] <Mavrik> neuro_sys, pretty much
[10:01:21 CEST] <fling> I'm reencoding mjpeg from a static camera to h264. How do I increase the time between keyframes properly? The image is mostly static.
[10:08:24 CEST] <furq> fling: -g and/or -x264opts keyint_min
[10:09:28 CEST] <furq> although -g should default to 250 which is a sensible value
[10:13:52 CEST] <JEEB> I recommend -x264params because I think that one actually just passes things to libx264 as-is
[10:14:02 CEST] <JEEB> while -x264opts had some extra internal parsing in it
[10:14:08 CEST] <JEEB> unless they are synonyms now
[10:14:19 CEST] <JEEB> or was it -x264-params... don't remember
[10:15:05 CEST] <JEEB> yeah, x264-params
[10:17:06 CEST] <JEEB> yeah, x264opts goes into an OPTS macro while x264-params passes the key-value pair straight into x264_param_parse
[10:17:10 CEST] <JEEB> which is libx264
[10:17:32 CEST] <JEEB> x264opts works with many things because they were manually defined in libavcodec's libx264.c :)
[10:17:33 CEST] <mrteatime> Sorry to dive right in, but anyone experienced with sending rgb32 rawvideo over udp who can help me?
[10:18:00 CEST] <mrteatime> Having some strange issues with playback and its related to the udp itself, as it playsback fine from a local file.
[10:19:28 CEST] <Mavrik> yuck
[10:19:31 CEST] <Mavrik> Why O.o
[10:20:36 CEST] <mrteatime> Why am i streaming rawvideo over udp? gonna use a usb 3.1 crossover cable to get a 10gbps stream from one pc to another, so i can offload the encoding to a seperate rig.
[10:22:08 CEST] <mrteatime> Problem is im doing local tests right now by playing back the file on the rig that is streaming it, and while i get the video feed and the color is correct, it spazzes the fuck out.
[10:22:41 CEST] <mrteatime> And by spazzes out i mean the frames look like they are just moving around freely, kinda like how you would imagine a fly sees tv.
[10:37:41 CEST] <Mavrik> mrteatime, that would probably be UDP at high bandwidths :)
[10:38:52 CEST] <mrteatime> Oh it is trust me, bandwidth requirement is about 3.7 GiBps but as i mentioned earlier, i have a 10 GiBps cap so bandwidth isnt a problem.
[10:39:54 CEST] <mrteatime> What im trying to figure out is why the playback isnt working properly, rawvideo alone works fine, udp works fine, but throw the two together and it just breaks.
[10:40:57 CEST] <mrteatime> Only thing i have to work off is that ffplay is giving me an invalid buffer size error, but for the life of me i cant find anyway to modify the buffer size let alone find the settings i need for it.
[10:41:24 CEST] <Mavrik> With UDP working correctly, do you mean that you verified if your packets are coming in in the proper order_
[10:41:25 CEST] <Mavrik> ?
[10:43:02 CEST] <mrteatime> The way i checked if udp was working is i did the same setup but with mpegts as the output instead of rawvideo
[10:43:56 CEST] <mrteatime> The mpegts came in just fine where as the rawvideo has the afforementioned problems (so hard to describe what it looks like haha)
[10:44:58 CEST] <mrteatime> if it helps these are my ffprompt settings, ffmpeg -f gdigrab -framerate 60 -offset_x 0 -offset_y 0 -video_size 1920x1080 -i desktop -f rawvideo -tune zerolatency udp:/127.0.0.1:1234, ffplay -video_size 1920x1080 -pixel_format bgra -framerate 60 -f rawvideo udp://127.0.0.1:1234
[11:05:34 CEST] <__jack__> hi, I've got an expert issue; Using an old self-compiled ffmpeg (probably from git, march 2015), the output ts has stream type 06 for audio; using ffmpeg 3.0, it now has stream type 87, and my stupid player won't handle it
[11:05:48 CEST] <__jack__> is there a way to add some kind of mapping or something ?
[11:06:16 CEST] <__jack__> (recompile is not really an option)
[11:08:23 CEST] <neuro_sys> Why are some functions postfixed with 2 such as avcodec_open2? Because once there was a avcodec_open, and kept backwards compatibility until it was deprecated and removed completely?
[11:08:52 CEST] <JEEB> yes
[11:08:58 CEST] <JEEB> first you get a warning
[11:09:03 CEST] <JEEB> when you build that is
[11:09:09 CEST] <JEEB> that you're using deprecated stuff
[11:09:15 CEST] <JEEB> then at some point it will stop compiling
[11:15:14 CEST] <__jack__> JEEB: no idea for my shitty issue ? :p
[11:15:32 CEST] <CoJaBo> The point it stops compiling is always when I try to compile it :/
[11:17:40 CEST] <CoJaBo> ffmpeg's errors are really terrible
[11:25:50 CEST] <__jack__> that id is define here: libavformat/mpegts.h:63, hum
[11:27:03 CEST] <BtbN> type 06 sounds entirely wrong for an audio stream.
[11:28:16 CEST] <mrteatime> just gonna re ask now that people are active, does anyone have experience with streaming rgb32 rawvideo over udp to ffplay
[11:29:07 CEST] <BtbN> don't stream raw video.
[11:29:12 CEST] <mrteatime> Everytime i do it ffplay is getting the stream but is reporting that the buffer size is incorrect and the frames are moving around rapidly.
[11:29:17 CEST] <mrteatime> Thats not an option btb
[11:29:40 CEST] <mrteatime> Im streaming uncompressed via a 10GiBps network connection to a seperate rig that will handle the encoding process.
[11:29:49 CEST] <mrteatime> Aka if its not raw i may aswell just not bother.
[11:30:09 CEST] <BtbN> Use a lossless codec, but streaming raw video frames won't work overly well.
[11:31:11 CEST] <mrteatime> The main reason i am going for raw is i want to avoid the encoding process all together for latency reasons on source machine, not so much for quality.
[11:31:16 CEST] <__jack__> BtbN: you're probably right, closed-source player powered, who knows how it works .. :)
[11:31:28 CEST] <mrteatime> Otherwise i wouldve gone lossless ages ago.
[11:31:53 CEST] <BtbN> raw video doesn't have timestamps or the slightest error correction.
[11:32:07 CEST] <BurnerGR> mrteatime, I had the same problem as you with udp
[11:32:10 CEST] <BtbN> Lose a single byte in transport and the stream is done for forever
[11:32:31 CEST] <__jack__> found a -mpegts_flags system_b that may do the job; Now looking at "how use -hls_ts_options with a boolean option"
[11:32:36 CEST] <__jack__> (if you've a clever tip :p)
[11:33:05 CEST] <BtbN> Use an IFrame-Only intermediade codec, like huffyuv.
[11:33:08 CEST] <BurnerGR> mrteatime, I ended up using pipe output to tcpdump, that works :)
[11:33:29 CEST] <BurnerGR> but the will be a point to point tcp connection
[11:35:13 CEST] <mrteatime_> The issue is i dont actually want timestamps in the feed.
[11:35:27 CEST] <mrteatime_> Because frankly im trying to avoid overhead, not add to it.
[11:36:41 CEST] <BurnerGR> mrteatime_, did you try with tcp instead?
[11:36:58 CEST] <BtbN> You definitely do want the overhead of some kind of container. As you already found out what happens without it.
[11:37:22 CEST] <BurnerGR> udp can arrive out of order, and might drop packets at random, even on LAN
[11:37:30 CEST] <mrteatime_> Btb no i didnt find out what happens without it because surprise surprise, i can actually play the rawvideo fine with no issues what so ever.
[11:37:40 CEST] <mrteatime_> Its only when i introduce it into a udp stream that errors occur
[11:37:52 CEST] <BtbN> mrteatime_, you just said you couldn't.
[11:37:52 CEST] <mrteatime_> thats what im trying to fix, not the codec
[11:38:03 CEST] <mrteatime_> no i said that i cant play it back via udp
[11:38:07 CEST] <BtbN> exactly
[11:38:17 CEST] <mrteatime_> if i record to a local rgb32 file and play that its no issue
[11:38:32 CEST] <BtbN> Of course, because you don't sent it over a network via udp.
[11:39:21 CEST] <BurnerGR> mrteatime_, did you try tcp?
[11:39:31 CEST] <mrteatime_> Trying it now burner
[11:39:37 CEST] <BurnerGR> ok :)
[11:39:49 CEST] <BurnerGR> netcat is your friend :)
[11:39:51 CEST] <BtbN> I'd guess tcp has a higher overhead than a lossless codec...
[11:40:00 CEST] <BurnerGR> indeed
[11:40:14 CEST] <mrteatime_> But btb just gotta mention, so you are saying that by sending a rawvideo feed to localhost, not over the network but in local testing, i am somehow magically putting the frames out of order
[11:40:35 CEST] <mrteatime_> despite having a theoretical infinite bandwith in my testing scenario and absolutely no packetloss
[11:40:45 CEST] <BtbN> It's UDP, even on localhost it can easily drop packets if you are sending massive amounts of data(like, raw video frames)
[11:41:52 CEST] <BtbN> A codec + container can compensate for that, a raw video stream is just irreversibly damaged if that happens.
[11:42:30 CEST] <BtbN> I wouldn't even trust TCP enough for a long running raw video stream.
[11:47:20 CEST] <BurnerGR> BtbN, I've run a setup like this for more than a year,  without problems, but my "long-running" streams are rarely more than 16 hours
[11:47:42 CEST] <BtbN> TCP is quite good at preventing this. But it's not 100% perfect.
[11:48:09 CEST] <BtbN> If something happens, it's usualy a bit flip or something like that, which doesn't break the stream.
[11:48:19 CEST] <BtbN> What breaks it is adding/removing bytes.
[11:49:47 CEST] <BurnerGR> I have a hard time finding a lossless codec which does not need a noicy server to keep up :/
[11:50:21 CEST] <BtbN> huffyuv, ffv1
[11:51:21 CEST] <Qann> Guys, I'm trying to use an external libav source I've compiled but I'm not sure whilst compiling if FFmpeg uses them or not. Does it? Or should I replace the libav* folders somehow
[12:25:41 CEST] <Bluez_> hi
[14:25:10 CEST] <fling> JEEB: which value is to use instead of 250 for a really static camera?
[15:40:33 CEST] <odinsbane> Can I tell the x264 encode to not use compression?
[15:41:48 CEST] <c_14> No
[15:41:53 CEST] <c_14> You can tell it to use lossless compression.
[15:42:00 CEST] <iive> odinsbane: x264 supports lossless compression, in case this is what you want
[15:42:30 CEST] <iive> if you don't want any compression, then don't use compressor :)
[15:44:33 CEST] <c_14> ffmpeg -codecs | grep -E ' .EV.+(Unc|raw)' <- should list most uncompressed/raw video codecs
[15:46:01 CEST] <odinsbane> Must mean lossless, or not to use another program to compress the video. A site is asking for H.264 encoded videos...do not compress the video.
[15:56:08 CEST] <DHE> maybe they mean don't try to ZIP it or something like that
[17:00:22 CEST] <Prelude2004c> hey guys
[17:00:53 CEST] <Prelude2004c> i keep getting errors like this and the encoding drops.. " [mpegts @ 0x36b7180] Non-monotonous DTS in output stream 0:0; previous: 38545191, current: 38545191; changing to 38545192. This may result in incorrect times
[17:00:54 CEST] <Prelude2004c> tamps in the output file. " .. anyone know any way around this ?
[17:01:02 CEST] <Prelude2004c> how do i create my own so its not dependant on a file or something
[17:01:27 CEST] <DHE> you doing this in C or with ffmpeg itself?
[17:04:46 CEST] <DHE> never mind, C doesn't "change" it
[17:15:24 CEST] <boegel> quick question: are there any backwards incompatible changes in FFmpeg 3.0 compared to 2.8.x?
[17:28:35 CEST] <jbreeden> I asked this question a couple weeks ago, figured I'd ask again to get a fresh pair of eyes on it: I have transport streams with embedded metadata (AVMEDIA_TYPE_DATA), and I'm trying to convert it to AVMEDIA_TYPE_SUBTITLE so that I can apply my custom parsing and export it as WebVTT. I've written the decoder, but it only works if I edit the TS demuxer to treat the data as subtitles from the get-go. Does FFmpeg provide a mechanism th
[17:29:04 CEST] <jbreeden>  be converted to a subtitle stream without mucking with FFmpeg's internals?
[17:49:01 CEST] <Prelude2004c> DHE ,, you talking to me ?  i am using ffmpeg
[17:54:40 CEST] <fling> JEEB: or should I tune the min value somehow?
[18:00:01 CEST] <neuro_sys> Is it possible to specify frame numbers rather than seconds (for cutting)?
[18:01:20 CEST] <furq> fling: scenecuts are only forced before the -g interval if there's a scenecut, which it sounds like there isn't
[18:01:48 CEST] <furq> and you can only seek to keyframes, so increasing -g much beyond 250 will cause problems with seeking
[18:02:07 CEST] <furq> i would just use the defaults, there aren't particularly great space savings to be had
[18:02:19 CEST] <furq> unless you're never going to seek this video
[18:03:49 CEST] <fling> furq: thanks! :>
[18:06:36 CEST] <fling> furq: what else could I tune to save some space with this fixed camera video?
[18:07:05 CEST] <fling> DHE: how exactly do I denoise?
[18:08:08 CEST] <DHE> fling: that question was answered in #x264
[18:08:32 CEST] <fling> Thanks.
[18:08:59 CEST] <furq> denoise, slower preset, higher crf, lower resolution, etc
[18:24:14 CEST] <vade> so im using XCode + guard malloc / guard edges /  / scribble - and I notice that it logs SSE should work - but then get an unnamed function for vmovdqu which is a AMX / simd extention used in libswscale - im wondering if im hitting a vector addition that guard malloc is unaware of in some inline assembly? Has anyone seen similar issues when debugging with ffmpeg?
[18:33:09 CEST] <fling> crf=23 with preset=veryslow is not giving me much space savings over mjpeg but takes a lot of time to reencode
[18:34:58 CEST] <fling> the resulting video takes ~70% of the original space
[19:02:58 CEST] <arbi> Anyone here used ffmpeg with cinelerra?
[19:03:37 CEST] <Elgul> no
[19:03:59 CEST] <arbi> You're answering for everyone? :-)
[19:04:07 CEST] <Elgul> no just for me
[19:04:09 CEST] <Elgul> ;__;
[19:04:45 CEST] <arbi> I just compiled latest ffmpeg and suddenly Cinelerra has problems rendering
[19:04:56 CEST] <Elgul> :/
[19:05:06 CEST] <Elgul> I never used cinelerra
[19:05:12 CEST] <arbi> ok
[19:05:16 CEST] <Elgul> is it even using ffmpeg?
[19:05:58 CEST] <arbi> there's a workaround for renderieng h264 which uses ffmpeg pipe
[19:06:09 CEST] <Elgul> nice
[19:06:57 CEST] <arbi> it worked just fine before.. why would newer ffmpeg have problems?
[19:07:39 CEST] <Elgul> I also had problems with ffmpeg when I used the self compiled version
[19:07:57 CEST] <Elgul> it always complained about vp9 and that I need to use -strict 2
[19:08:04 CEST] <Elgul> with the repo versions this never happend to me
[19:08:30 CEST] <arbi> ffmpeg itself seems to work ok.. it's cinelerra that wont render
[19:08:49 CEST] <Elgul> I also had a rendering problem 2 weeks ago while using kdenlive
[19:08:54 CEST] <arbi> maybe it can't locate ffmpeg I don't know
[19:09:03 CEST] <arbi> hmm
[19:09:09 CEST] <Elgul> and I also thought it would be a ffmpeg problem
[19:09:11 CEST] <Elgul> but it wasnt
[19:09:21 CEST] <furq> arbi: does it use the ffmpeg libs or the binary
[19:09:27 CEST] <Elgul> in my case I just used the blur effect too much
[19:09:57 CEST] <arbi> It pipes like this
[19:09:58 CEST] <arbi>  ffmpeg -f yuv4mpegpipe -i - -y -vcodec libx264 -crf 21 -preset medium %
[19:10:59 CEST] <furq> and what error do you get
[19:12:08 CEST] <arbi> Can't copy the darn error message.
[19:12:30 CEST] <arbi> int YUVStream::write_frame(uint8 t**): write_frame() failed: system error (failed read/write)
[19:12:41 CEST] <arbi> That's the error
[19:12:59 CEST] <Elgul> read/write could this be a permission thing?
[19:13:02 CEST] <arbi> Problem is that you get the same error if you don't have ffmpeg installed
[19:13:32 CEST] <arbi> It's not permission.. same user can run it in the terminal
[19:14:22 CEST] <arbi> The only change that I've done is to compile new ffmpeg. Which is globally accessable
[19:14:30 CEST] <furq> cinelerra's website is really remarkably bad
[19:14:49 CEST] <arbi> :)
[19:15:34 CEST] <arbi> Well Cinelerra is pretty good video editor..
[19:15:52 CEST] <Elgul> it could also be a PATH thing
[19:15:56 CEST] <Elgul> have you checked that?
[19:16:14 CEST] <arbi> Elgul: What should I check?
[19:16:23 CEST] <Elgul> if ffmpeg is in your path
[19:16:25 CEST] <furq> all i want to know is if it depends on libavcodec et al but this website is totally useless and it's not in any repos
[19:16:55 CEST] <arbi> which site are you looking?
[19:16:59 CEST] <furq> cinelerra.org
[19:17:04 CEST] <arbi> I'm actually using Cinelerra-cv
[19:17:55 CEST] <arbi> Elgul: The path to new version is in $PATH
[19:17:59 CEST] <furq> oh this website is less useless
[19:18:05 CEST] <arbi> hehe
[19:19:07 CEST] <furq> never mind, it looks like it ships with libavxyz so presumably they're static linked
[19:19:13 CEST] <furq> which makes sense i guess
[19:19:53 CEST] <arbi> :-/
[19:23:47 CEST] <cowai> $ ffmpeg -loglevel error -loop 1 -i static.jpg -f lavfi -i aevalsrc=0 -t 1  -shortest -strict experimental -c:a aac -b:a 256000 -ar 48000 -ac 2 -c:v libx264 -profile:v high -level 4.0  -b:v 3000000 -r 30 -g 30 -s 1920x1080 -pix_fmt yuv420p -f matroska -y "/tmp/test.mkv" && ffprobe -hide_banner /tmp/test.mkv 2>&1 |grep Duration
[19:23:47 CEST] <cowai>   Duration: 00:00:01.02
[19:24:36 CEST] <cowai> I need to generate a mkv file with empty audio track and a static image for video that last exactly 1 second or 30 frames
[19:25:04 CEST] <furq> cowai: -frames:v 30
[19:25:35 CEST] <cowai> still 1.02 with that option
[19:26:02 CEST] <cowai> if I remove "-t 1" I get 1.20 instead of 1.02
[19:26:50 CEST] <cowai> I have tried ffmpeg 2.8 and 3.0.
[19:26:54 CEST] <cowai> same result
[19:28:31 CEST] <cowai> If I replace the audio params with "-an" and keep -t 1 I get 1.00 second
[19:29:09 CEST] <cowai> But I need there to be a silent audio track because I am concatting these files with some videos files that has audio.
[19:30:54 CEST] <cowai> @furq: any ideas?
[19:31:00 CEST] <furq> what happens if you remove -shortest
[19:31:30 CEST] <cowai> 1.02
[19:32:28 CEST] <cowai> if I use libvorbis instead of aac, I get 1.0
[19:32:33 CEST] <cowai> *1.00
[19:34:49 CEST] <brontosaurusrex> is opus any less supported (in webm with vp9 video) than vorbis this days?
[19:34:59 CEST] <furq> i guess it's something to do with aac frame duration then
[19:35:06 CEST] <brontosaurusrex> browser playback I mean.
[19:35:27 CEST] <furq> i don't think 1000 is divisible by the frame duration at 48khz, but i'm not an expert on aac
[19:35:56 CEST] <furq> brontosaurusrex: not on desktop
[19:36:02 CEST] <drv> http://caniuse.com/#feat=opus
[19:36:06 CEST] <furq> yeah
[19:36:10 CEST] <furq> http://caniuse.com/#search=vorbis
[19:36:18 CEST] <furq> only difference is the android browsers
[19:36:47 CEST] <furq> edge 14 will have opus and not vorbis, though, so it cancels out
[19:37:05 CEST] <cowai> furq: you may be right, when I use 25 fps it works
[19:37:47 CEST] <furq> by works do you mean 30 frames or 1 second
[19:37:52 CEST] <brontosaurusrex> thanks
[19:37:57 CEST] <cowai> 1 second
[19:38:17 CEST] <cowai> is it not possible to get a 1 second clip with aac ?
[19:38:18 CEST] <furq> the video framerate wouldn't affect that if it is to do with aac frame duration
[19:40:07 CEST] <furq> but yeah afaik at 48khz each aac frame is 21.33ms
[19:40:50 CEST] <brontosaurusrex> hmm, so android has crappy opus support for some reason?
[19:41:04 CEST] <cowai> is there another samping rate I can use that would fit?
[19:54:58 CEST] <brontosaurusrex> actually 94% for h.264 vs 73% for vp9 (localy), probably not really worth bothering.
[19:56:27 CEST] <Prelude2004c> hey guys.. sorry to be a bother.. i still can't figure out this..
[19:56:28 CEST] <Prelude2004c> " [mpegts @ 0x36b7180] Non-monotonous DTS in output stream 0:0; previous: 38545191, current: 38545191; changing to 38545192. This may result in incorrect times
[19:56:43 CEST] <Prelude2004c> randomly my files show that... and i don't know what to do about it.. it crashes up the whole thing :(
[19:56:49 CEST] <Prelude2004c> how do i ignore those errors and keep going
[19:57:05 CEST] <Prelude2004c> can i not generate my own timestamps or DTS or something
[19:57:17 CEST] <Prelude2004c> so it its fresh and does not care too much about the source issue
[19:57:31 CEST] <Prelude2004c> when i play the file on VLC , it works fine :(. when i try to transcode it using ffmpeg.. it acts up
[20:06:04 CEST] <vade> im concerned im misunderstanding something or configuring my encoders video stream time base and video streams codec context time base incorrectly, as im getting way higher frame rates for some video than I should be. Does this look correct? https://gist.github.com/vade/4a1084e697c0a3cb33ffd6ce6fe47f6e
[20:06:05 CEST] <vade> my encoder has a reference to the decoder to get its video stream info to match
[20:06:34 CEST] <vade> but im wondering if im setting up ticks_per_frame incorrectly, or should not be poking at the codec context struct directtly. Something is definitely wron
[20:06:35 CEST] <vade> g
[20:08:56 CEST] <computer> people i am having trouble converting a flash video, can someone help me? please?
[20:16:14 CEST] <computer> ffmpeg is changing my screen and zooming greatly, how can i avoid that?
[20:18:11 CEST] <computer> i am getting this:
[20:18:14 CEST] <computer> Input stream #0:0 frame changed from size:58x62 fmt:argb to size:42x47 fmt:argb
[20:22:38 CEST] <neuro_sys> Could anyone point out the link to between() in documentation?
[20:22:54 CEST] <furq> neuro_sys: http://ffmpeg.org/ffmpeg-utils.html#Expression-Evaluation
[20:23:05 CEST] <computer> ?
[20:24:19 CEST] <neuro_sys> furq: thanks
[20:24:25 CEST] <computer> anyone understands my bug?
[20:25:13 CEST] <neuro_sys> so to apply a filter between a specific interval I should be using enable=between(t, s, e). However can it operate on the timestamp format? (still reading.)
[20:26:10 CEST] <furq> neuro_sys: https://ffmpeg.org/ffmpeg-filters.html#Timeline-editing
[20:26:42 CEST] <neuro_sys> furq: It's only seconds precision? Actually I needed milliseconds precision.
[20:27:46 CEST] <furq> looks like it
[20:27:57 CEST] <neuro_sys> oh bummer...
[20:28:06 CEST] <furq> i've never needed ms precision independent of framerate, so i'm not totally sure
[20:28:30 CEST] <neuro_sys> I'd rather work with frame numbers, but apparently ffmpeg using timestamp for such needs.
[20:28:52 CEST] <furq> replace t with n to use frame numbers
[20:29:13 CEST] <neuro_sys> furq: oh, right
[20:29:22 CEST] <computer> guys what is the best format to convert .swf files to?
[20:29:54 CEST] <furq> computer: if it's not being decoded properly then it makes no difference which output format you use
[20:30:26 CEST] <neuro_sys> furq: FWIW, I have a set of images (generated by opengl), that I need to overlay between certain intervals on a video.
[20:30:38 CEST] <computer> furq: what do you suggest i do? i have been reading the documentation today, but did not understand how to fix my issue
[20:31:09 CEST] <computer> there is nothing wrong with the file, it plays perfectly. but i cannot convert it properly
[20:31:20 CEST] <furq> i have no idea how good ffmpeg's support for swf is
[20:31:26 CEST] <furq> i can't imagine it's possible for it to be complete
[20:31:43 CEST] <furq> there's no correct way to convert an interactive swf, for example
[20:32:37 CEST] <computer> furq: hmmm, the video has a mouse moving, but no interaction (buttons, clicks etc...) and effmpeg is only converting the mouse movement
[20:35:51 CEST] <neuro_sys> Am I missing something here (testing): -filter_complex "[0:v][1:v]overlay:enable='between(t, 1, 1)'[out]" and I get
[20:35:54 CEST] <neuro_sys> No such filter: 'overlay:enable'
[20:36:24 CEST] <furq> overlay=enable=
[20:36:35 CEST] <neuro_sys> oh yes
[20:37:12 CEST] <neuro_sys> between(t, 1, 2) but anyways, thanks
[20:38:12 CEST] <computer> furq: is there a way to by pass this and just convert the whole file, not the mouse movement?
[20:53:44 CEST] <yongyung> I want to convert a video to a most easily decodeable, lossless format. So essentially uncompressed. pix_fmt is yuv420, but what would I use as codec?
[20:54:18 CEST] <neuro_sys> Let's say I have N number of input streams (each one a set of images fed from stdin in the form of %4d.png). Can I apply a distinct overlay filter using enable=between with each of the input streams for different intervals? I'm thinking along the lines of ffmpeg -i base -i input1 -i input2  -filter_complex "[0:v][1:v]overlay=enable='between(n,1,2)'[out1]" -filter_complex
[20:54:23 CEST] <neuro_sys> "[out1][2:v]overlay=enable='between(n,3,4')[out2]" -map '[out2]' output.mp4
[20:55:12 CEST] <computer> can someone help me covert my file? it is only 1 minute long
[20:55:45 CEST] <furq> neuro_sys: you can only call filter_complex once
[20:56:07 CEST] <furq> separate them with ;
[20:56:32 CEST] <neuro_sys> furq: It worked!
[20:56:35 CEST] <neuro_sys> thanks again
[20:56:49 CEST] <llogan> yongyung: -c:v rawvideo if you want raw video
[20:56:59 CEST] <neuro_sys> I still should take the time to use libav* with C though... (if only the deadline wasn't a couple days away)
[21:00:09 CEST] <furq> yongyung: probably ffv1
[21:02:03 CEST] <yongyung> rawvideo seems fine
[21:02:35 CEST] <furq> if you have enough space and disk speed for rawvideo then sure
[21:03:12 CEST] <yongyung> It's just for testing purposes^^
[21:13:33 CEST] <arbi> furq: If you're interested the write fail problem with ffmpeg was caused by a space in the filename
[21:22:34 CEST] <Qann> Guys, I'm trying to use an external libav source I've compiled but I'm not sure whilst compiling if FFmpeg uses them or not. Does it? Or should I replace the libav* folders somehow
[22:29:40 CEST] <brontosaurusrex> A nice CRF + fast + multithreaded VP9 example?
[22:43:30 CEST] <neuro_sys> "[image2 @ 0x5ec4260] Thread message queue blocking; consider raising the thread_queue_size option (current value: 8)" Is this something to be concerned?
[22:44:44 CEST] <durandal_1707> how many threads you use?
[22:45:08 CEST] <neuro_sys> The message says 8, but I'm not giving any specific parameters for it
[22:45:39 CEST] <durandal_1707> how many cores/cpus?
[23:31:45 CEST] <Prelude2004c> hey, can anyone help ? [mpegts @ 0x36b7180] Non-monotonous DTS in output stream 0:0; previous: 38545191, current: 38545191; changing to 38545192. This may result in incorrect times
[23:42:21 CEST] <vade> i understand that when one calls avformat_write_header a streams time_base might be altered - what prompts a drastic change from requesting 2997 to getting 11984 ?
[23:47:32 CEST] <neuro_sys> when I feed a video stream as %4d.png format, it doesn't work, while for indivial pngs it works, any ideas?
[23:55:24 CEST] <Mavrik> vade, hrmf, it shouldn't really.
[23:55:29 CEST] <Mavrik> Which muxer?
[23:55:56 CEST] <vade> hi Mavrik - im using MP4
[23:56:59 CEST] <Mavrik> hm, I don't see the source touching that field
[23:57:00 CEST] <vade> Mavrik: I log my settings . I build my video stream and video streams -> codec via introspection on the source
[23:57:05 CEST] <vade> https://gist.github.com/vade/4acd447dab0dd9b307b7ffe49625f179
[23:57:18 CEST] <vade> let me share how I set up my videoStream
[23:58:05 CEST] <vade> Mavrik: https://gist.github.com/vade/2d3853f55de30e9fa8c878f54d8dafaa
[23:59:28 CEST] <vade> oh, that snipped was from the wrong video (the output from my app) Its weird. I cant seem to get my app to return the right ffprobe output for fps, tbn, tbc,
[00:00:00 CEST] --- Tue May 10 2016