[Ffmpeg-devel-irc] ffmpeg.log.20150920

[burek]

[02:49:46 CEST] <fred1807> can I use ffmpeg to blend 2 videos?   one full screen at the backround, and a small square size centered on top ?
[02:58:45 CEST] <Rudde> Hello
[03:03:55 CEST] <DHE> fred1807: based on a reading of ffmpeg-filter, it can technically be done but depends on the overlay image having an alpha channel to do the blending
[03:04:31 CEST] <fred1807> nice
[03:05:37 CEST] <Rudde> So I installed ffmpeg froum source on debian to use with pHash
[03:05:54 CEST] <Rudde> but after I installed it I can run ./configure on pHash it says ffmpeg is not installed
[03:05:57 CEST] <Rudde> when it is
[05:54:29 CEST] <hotwings> is there a way to get ffprobe to print only specific info? i only want to extract the number of audio channels in the first audio stream. i dont want any headers or other stream info. i tried "ffprobe -hide_banner -show_streams -select_streams a:0 file.mkv |grep "^channels" but it still shows this stream info stuff
[06:59:06 CEST] <relaxed> hotwings: add 2>&1 after file.mkv
[07:49:59 CEST] <hotwings> thanks relaxed, ill try that
[07:51:23 CEST] <hotwings> that worked, thanks again!
[09:34:57 CEST] <relaxed> hotwings: ff* prints console output to stderr and stdout, 2>&1 redirects stderr to stdout. Now you don't need -hide_banner
[11:30:01 CEST] <luc4> Hello! Is there a way with ffmpeg to dump a file with the content of the frames that the function av_read_frame reads? I need this for debugging purposes.
[11:52:53 CEST] <xerox> Hi. I have H.264 video encoded with (what I believe to be) parameters suitable for fast-moving scenes. These files are quite large. The video is actually comprised of very still frames with small parts only moving once in a while. Another important aspect is that I don't want them to get blurry at all. Where can I find some information to get started with such a re-encoding project? Hope it's okay to ask her
[11:52:59 CEST] <xerox> e (:
[11:55:30 CEST] <Mavrik> Hmm.
[12:03:21 CEST] <xerox> The reason why I don't want them to get blurry is because there is a lot of text in very small size.
[12:03:49 CEST] <Mavrik> I guess using crf with one of the slow presets would work.
[12:04:01 CEST] <Mavrik> But I don't know what "large" means to you and how much improvement do you expect.
[12:04:09 CEST] <Mavrik> Maybe even -tune stillimage
[12:05:12 CEST] <xerox> Here's a sample mediainfo http://pastebin.com/raw.php?i=NhpFPT0z
[12:06:22 CEST] <Mavrik> 3Mbit for 1080p is already a very good compression.
[12:06:26 CEST] <DHE> xerox: sounds like you might want to set qpmax to something smaller. maybe in the low 30s?
[12:07:22 CEST] <xerox> Mavrik: I think so too, but this is really not your general case video, it is basically like a screen capture of a desktop with not much going on at all.
[12:08:02 CEST] <xerox> OK (1) crf with slow preset; (2) -tune stillimage; (3) qpmax to something small
[12:08:32 CEST] <DHE> ffmpeg will call it -qmax I think... I mostly use x264 parameters directly
[12:09:02 CEST] <DHE> the preset will mostly affect the bitrate with -crf. actual image quality will have minimal impact
[12:10:41 CEST] <xerox> Oh right I also need a nudge in the right direction as far as practical ways to do this, can you recommend a simple tutorial I can get inspiration from to do such a re-encode?
[12:11:16 CEST] <xerox> I have both ffmpeg and x264 installed (on osx, via brew.)
[12:11:41 CEST] <DHE> honestly you just need to read what the options do and experiment a bit.
[12:12:20 CEST] <xerox> I found the wikibook on x264 that explains those, I'll go through them :D
[12:12:49 CEST] <DHE> there's a lot, and some of them are not obvious (like the bframe options)
[12:14:36 CEST] <ewew> rawvideo :)
[12:21:04 CEST] <luc4> Hello! Im trying to link my application to ffmpeg 2.8 from ffmpeg 2.7.2. What I see is that it works no more with 2.8. Unfortunately still I cant see what is wrong. Any of you who has an advice on how I could debug this?
[12:21:50 CEST] <ewew> @luc4 Use ffmpeg alone on command line and add debug flag.
[12:22:03 CEST] <luc4> ewew: ffplay alone works properly.
[12:22:53 CEST] <luc4> ewew: I can only reproduce the problem with one video file.
[12:22:56 CEST] <ewew> @luc4 There might be some syntax changes. Review your ffmpeg parameters.
[12:23:51 CEST] <luc4> ewew: I was pointed to the changelog, but I couldnt find any difference& https://github.com/FFmpeg/FFmpeg/blob/master/doc/APIchanges#L41
[12:24:15 CEST] <luc4> ewew: I see some macro names have changed (I changed those), and the rest is mostly something added...
[12:24:29 CEST] <luc4> ewew: and not using vaapi
[12:24:36 CEST] <luc4> ewew: I am just demuxing
[12:25:39 CEST] <luc4> ewew: Ive also done another check: I dumped to file all the encoded packets coming from ffmpeg 2.7.2 and 2.8, compared the hexdump with meld and& identical.
[12:26:05 CEST] <luc4> ewew: still if I build my app to 2.7.2 everything is ok, 2.8 is not.
[12:26:19 CEST] <luc4> ewew: really no idea how to debug this...
[12:26:43 CEST] <ewew> wait i will check the documentation.
[12:27:16 CEST] <luc4> ewew: I could even provide the video but not sure if it can be of any help& it really seems an issue of my app, but I clearly see it is ok with 2.7.2 and ok with most files with 2.8...
[12:27:53 CEST] <luc4> ewew: still packets coming from ffmpeg are perfectly equal, do no idea at all of what is happening :-)
[12:28:17 CEST] <ewew> https://ffmpeg.org/ffmpeg-all.html#Generic-options
[12:28:17 CEST] <ewew> parameter -loglevel debug.
[12:29:31 CEST] <luc4> ewew: ok, that is the param of the ffmpeg command, but Im using libavformat.
[12:33:10 CEST] <ewew> Exactly what tool you are using ? ffmpeg or any other ?
[12:34:24 CEST] <luc4> ewew: Im using libavformat to read encoded frames, then I feed those frames to hardware encoders and render using hw accel.
[12:34:59 CEST] <ewew> How is ffmpeg involved in this ?
[12:35:13 CEST] <luc4> ewew: link my code to 2.7.2 -> everything is great. link the same exact code to 2.8 -> on that file I see nothing at all.
[12:35:40 CEST] <luc4> ewew: I am the first to say I have no idea& but simply changing ffmpeg version changes the behavior...
[12:36:26 CEST] <luc4> unfortuantely I see there are many many changes from 2.7.2 to 2.8
[12:37:03 CEST] <luc4> also, if I simply reencode or remux somehow that file, my app with 2.8 works again.
[12:37:51 CEST] <ewew> First thing is to use ffmpeg alone with debug, the we can get somewhere. Try the #libav IRC room.
[12:38:51 CEST] <luc4> ewew: ah ok, wrong room, sorry.
[13:02:38 CEST] <xdbx> hey guys... any idea why ffmpeg depends on smbclient ?
[13:02:46 CEST] <xdbx> doesnt make sense for me
[13:11:29 CEST] <xdbx> cant get it to configure cuz of that
[13:11:34 CEST] <xdbx> anyone here ?
[13:14:04 CEST] <iive> patience...
[13:14:08 CEST] <dongs> gah
[13:14:10 CEST] <iive> what is the exact error?
[13:14:24 CEST] <dongs> PTS -442721857133401536, next:13242125480 invalid dropping st:1
[13:14:25 CEST] <dongs> DTS -1660206964250255104, next:13241348141 st:0 invalid dropping
[13:14:27 CEST] <dongs> im getting this shit
[13:14:29 CEST] <dongs> over and over
[13:14:33 CEST] <Mavrik> xdbx, it doesn't.
[13:14:45 CEST] <dongs> files are .mp4 from hdcam
[13:16:45 CEST] <xdbx> Mavrik, so why do I get "ERROR: smbclient not found" then ? =/
[13:16:56 CEST] <Mavrik> No idea.
[13:17:03 CEST] <dongs> https://trac.ffmpeg.org/ticket/3598 i guess this is kinda related
[13:17:12 CEST] <dongs> i start getting that dts/pts spam
[13:17:14 CEST] <dongs> from 2nd file
[13:19:06 CEST] <dongs> hm looks like i ahve to intermediate them to .ts first
[13:29:44 CEST] <DHE> you can disable libsmbclient with configure options, no?
[13:30:02 CEST] <fritsch> he is already gone and got his answer
[13:30:08 CEST] <fritsch> in the dev channel
[16:33:44 CEST] <joeschmo123> hello
[16:34:10 CEST] <joeschmo123> i have a TV tuner device
[16:34:19 CEST] <joeschmo123> is FFMPEG able to use that as an input device?
[16:34:35 CEST] <joeschmo123> if so, how do i tune it / set it to the proper channel?
[16:37:44 CEST] <voicesl> depends on the tuner device
[16:38:07 CEST] <iive> joeschmo123: ffmpeg support v4l2 input
[16:38:11 CEST] <joeschmo123> right now i can view the TV using windows media center, then i use ffmpeg to stream my whole screen
[16:38:24 CEST] <voicesl> I just recently did a small side project to make ffmpeg with with a HD HomeRun Prime via UDP
[16:39:22 CEST] <joeschmo123> this is the tuner device i have: http://www.pctvsystems.com/Products/ProductsNorthAmerica/HybridproductsUSA/PCTVHDProStick/tabid/169/language/en-US/Default.aspx
[16:40:16 CEST] <iive> usb dongle?
[16:40:32 CEST] <joeschmo123> yes.  antenna attaches to it.
[16:40:48 CEST] <iive> analog or digital capture?
[16:40:53 CEST] <joeschmo123> digital
[16:41:09 CEST] <iive> ok, so the keyword is dvb
[16:41:20 CEST] <joeschmo123> well, it supports both but i use it for digital because thats the only thing broadcasted here
[16:41:33 CEST] <iive> v4l2 is for analog.
[16:42:17 CEST] <iive> Yeh, i do have pctv 510e, it's quadro, composite, analog, dvb-t/c . unfortunately the analog part is not supported (afaik)
[16:43:01 CEST] <Mavrik> not sure what analog/digital has to do with capture interace.
[16:43:06 CEST] <Mavrik> *interface.
[16:43:23 CEST] <Mavrik> joeschmo123, figure out which device API your OS + device uses, then use that one to capture the input
[16:43:35 CEST] <Mavrik> It's really hard to tell with random cheap crap without actually having it.
[16:46:00 CEST] <joeschmo123> Mavrik: how can i tell?
[16:46:19 CEST] <joeschmo123> if this card doesn't work i may want to buy one that is supported with ffmpeg
[16:46:33 CEST] <ewew> What os you are using ? linux,osx,windows ?
[16:46:42 CEST] <joeschmo123> windows
[16:46:53 CEST] <Mavrik> Then it has to expose the interface via directshow
[16:46:59 CEST] <Mavrik> If it doesn't, you're pretty much SoL
[16:48:42 CEST] <joeschmo123> ok, it shows up as a directshow device
[16:48:43 CEST] <joeschmo123> http://pastebin.com/EGtPTZ5f
[16:48:51 CEST] <joeschmo123> DiBcom BDA Analog Capture (Dev1 Path2)
[16:54:11 CEST] <joeschmo123> i get a blue video when i try this command:
[16:54:13 CEST] <joeschmo123> ffmpeg -f dshow -i video="DiBcom BDA Analog Capture (Dev1 Path2)" out.mp4
[16:55:25 CEST] <Mavrik> see https://www.ffmpeg.org/ffmpeg-devices.html#dshow
[16:55:39 CEST] <Mavrik> try using -list_devices or -list_formats to see if you can enable the right input
[16:55:53 CEST] <Mavrik> you may have to pass video_device_number or some such
[16:56:16 CEST] <joeschmo123> thanks Mavrik
[16:56:18 CEST] <joeschmo123> let me try that
[16:57:32 CEST] <Mavrik> Usually those devices expose number of "inputs" and tend to crap out unless you hit right preferences.
[16:57:49 CEST] <Mavrik> But blue video is good, it means it's talking to you.
[17:05:38 CEST] <joeschmo123> i got a list of DirectShow video device options
[17:06:02 CEST] <joeschmo123> Crossbar Input pin 0 - "Video Tuner" related input pin: 3
[17:07:01 CEST] <joeschmo123> there must be some way to tell the tuner to change frequencies
[17:11:53 CEST] <joeschmo123> Windows Media Center can interface with the TV Tuner fine... i wonder if there's a way to debug that and find out how it's communicating with the device and try to get ffmpeg to do the same thing
[17:13:28 CEST] <KraftyWA> I'm creating a mosaic of videos. It's very CPU intensive to combine 4 videos. Each video is ~4500 bitrate, 1280x720. I'm shrinking them to 1/2 size and then placing them in corners. I was looking at the output's bitrate. It's roughly equal to all the input video bitrates added together (16000). That seems really high. Could someone explain to me why that is? I'm using '-c:v libx264 -crf 23 -r 29.97 -pix_fmt yuv420p'. Any tips for sp
[17:14:03 CEST] <joeschmo123> KraftyWA: you can set the bitrate for the output video
[17:14:21 CEST] <joeschmo123> e.g. -b:v 5000k
[17:14:52 CEST] <KraftyWA> But why does ffmpeg want such a high bitrate? Wouldn't I be losing lots of quality to force it back down to what the video input bitrates are?
[17:15:35 CEST] <joeschmo123> i'm not sure, but i think by default ffmpeg uses whatever bitrate it needs to keep 100% quality
[17:16:19 CEST] <joeschmo123> it will still look good at lower bitrates
[17:16:31 CEST] <KraftyWA> Yeah, that's what I figured. I just expected to see the bitrate be a mean of the inputs, not a sum.
[17:17:52 CEST] <KraftyWA> I wouldn't really mind, but right now it's taking 3 seconds to encode 1s. Not great if you're trying to watch while it's encoding the streams.
[17:24:08 CEST] <DHE> when it finishes x264 will print stats. you can press 'q' to make it finish now (obviously with an incomplete file). the numbers sometimes give clues
[17:28:19 CEST] <KraftyWA> My mocias shows 13377 kb/s. I assume that's the average bitrate. Seems so high.
[19:16:02 CEST] <user3> how can I catenate the files 1.flv 2.flv 3.flv ... 20.flv
[19:29:25 CEST] <votz> user3: Were they all encoded with the same settings?
[19:44:20 CEST] <user3> yes
[19:46:16 CEST] <votz> user3: Then standard approaches should apply
[19:46:44 CEST] <votz> user3: See https://trac.ffmpeg.org/wiki/Concatenate
[20:22:32 CEST] <kyleogrg> hey
[20:23:02 CEST] <kyleogrg> i want to do a test with x264 where i encode black video and highly-complex video at crf 18 and compare the bitrates.
[20:23:16 CEST] <kyleogrg> what type of video is hardest to encode and uses most bitrate?  any example?
[20:23:36 CEST] <kyleogrg> Would it be some kind of colorful static?
[20:24:24 CEST] <RobotsOnDrugs> probably
[20:24:24 CEST] <RobotsOnDrugs> what is the point of this?
[20:24:42 CEST] <waressearcher2> when I concatenate videos using: "ffmpeg -f concat -i list.txt -c copy -y -f avi out.avi" is it possible to have it read first file from 20 second position ? or should I first do it separately and then concatenate ?
[20:27:35 CEST] <kyleogrg> RobotsOnDrugs: I want to test the "relative compressability" of two videos.  I was going to do CRF 18 and preset ultrafast and see how much higher in % the "complex" bitrate is than the "simple" bitrate.
[20:28:04 CEST] <kyleogrg> Then I would slow down the preset step by step and again compare the bitrate % difference.
[20:28:52 CEST] <kyleogrg> I want to see how fast the preset can be while getting an even amount of compression across different types of video
[20:28:57 CEST] <RobotsOnDrugs> the difference is going to be huge
[20:29:48 CEST] <kyleogrg> yeah i figured
[20:30:31 CEST] <RobotsOnDrugs> they are two extreme ends, and most video you'll encounter will not be near either one in terms of compressibility
[20:31:12 CEST] <kyleogrg> ok, so let's say it's two real-life videos, but they are still very different in quality
[20:31:56 CEST] <kyleogrg> i would expect that the slower the preset, the bigger the bitrate difference.  is that right?
[20:33:15 CEST] <RobotsOnDrugs> in general, yes (to a point)
[20:33:23 CEST] <kyleogrg> so if video2 is 2x larger (in file size) than video1 at ultrafast, would it also be 2x larger at veryslow?
[20:33:40 CEST] <kyleogrg> right... so i'm wondering, what is the point where you get diminishing returns?
[20:34:24 CEST] <Mavrik> A perfect white noise won't make any difference in presets.
[20:34:28 CEST] <RobotsOnDrugs> depends on the video
[20:34:29 CEST] <Mavrik> Because there's no pattern to be found.
[20:35:06 CEST] <RobotsOnDrugs> uncompressible stuff will keep a high bitrate
[20:35:22 CEST] <kyleogrg> well, here's why i want to do this.  i'm encoding a batch a videos, and i want them to have a consistent quality and have a predictable batch size.
[20:35:29 CEST] <RobotsOnDrugs> very compressible stuff will get better, but only to a point
[20:36:03 CEST] <kyleogrg> so i was going to do a fast first pass at CRF 18 and get the bitrate
[20:36:11 CEST] <RobotsOnDrugs> that's going to be very difficult
[20:36:33 CEST] <kyleogrg> then later i could compare the bitrate reports and assign a bitrate for each video given the target batch size
[20:36:37 CEST] <RobotsOnDrugs> it is possible with a lot of effort, but not really practical
[20:40:13 CEST] <kyleogrg_> back :/
[20:40:18 CEST] <kyleogrg_> i was disconnected
[20:40:45 CEST] <waressearcher2> any thoughts ?
[20:41:18 CEST] <kyleogrg_> so basically i'm wondering, if i did this unusual method, what preset should be used for the first pass to get an accurate "complexity rating"?
[21:43:55 CEST] <fritsch> https://twitter.com/gsuberland/status/529046083134324736/photo/1
[21:47:10 CEST] <waressearcher2> how to add black rectangles on top and bottom ?
[21:48:28 CEST] <waressearcher2> I know there is filter "drawbox" but I want to increase height of the video by adding black rectangles, it is 720 now I want to increase to 960 without stretching original video
[21:49:01 CEST] <relaxed> waressearcher2: pad filter
[21:50:11 CEST] <waressearcher2> found it -vf "pad=1920:1080:0:20,scale=1280:720"
[23:13:45 CEST] <Mavrik> hmm, libx265 with ffmpeg doesn't set vbv parameters.
[23:42:18 CEST] <JEEB> Mavrik: if you mean through the so-called "generic" options then yes. I found that out with quite a few options some time ago (including -g). Use x265-params and set libx265 API key-value pairs
[23:42:38 CEST] <Mavrik> Yeah, I was looking at the source.
[23:43:02 CEST] <Mavrik> The 2.8 brings more in but for some reason vbv aren't in yet and was just wondering if they're ever functional in the libx265 itself.
[23:44:01 CEST] <JEEB> libx265 does have VBV and it seemed to work "OK". you just have to set the options through API key-value pairs
[23:44:07 CEST] <JEEB> through x265-params
[23:45:22 CEST] <Mavrik> Ack.
[23:45:24 CEST] <Mavrik> Thanks.
[00:00:00 CEST] --- Mon Sep 21 2015