[Ffmpeg-devel-irc] ffmpeg.log.20150827

burek burek021 at gmail.com
Fri Aug 28 02:05:01 CEST 2015


[00:49:35 CEST] <Phr33d0m> hi, I'm trying to output screenshots from a mkv file yet I also want the subtitles (probably ass or ssa codec) to be printed as well on the screenshot
[00:50:09 CEST] <Phr33d0m> I've tried `ffmpeg -ss 22:35 -i somevideo.mkv -t 51 -s 720x480 -map 0 -map -0:a -scodec ssa -f image2 /tmp/%03d.png` changing ssa for ass without success
[02:07:07 CEST] <guest4377> does ffmpeg enforce some limit on png dimensions?
[02:07:27 CEST] <guest4377> I'm trying to encode a 65536x65536 image, but it says "[png @ 03f29060] [IMGUTILS @ 0022f334] Picture size 65536x65536 is invalid"
[02:20:21 CEST] <kepstin> lets see... that's hitting the limit in av_image_check_size
[02:20:51 CEST] <kepstin> which is approximately height * width must be less than 1/8th the value of INT_MAX
[02:21:12 CEST] <kepstin> i'm guessing it's a protection against overly large memory allocations
[02:23:03 CEST] <kepstin> you do realize that a 65536x65536 image would require 16gb of ram to store decoded, right?
[02:23:18 CEST] <kepstin> (assuming rgba or rgb0)
[02:25:44 CEST] <tommd> Some people do have 1TB machines, it doesn't seem entirely unreasonable that someone might have a large use case.
[02:26:13 CEST] <kepstin> I suspect that increasing the max image size in ffmpeg would require very careful auditing of a lot of code paths to switch to 64bit values.
[02:27:22 CEST] <Plorkyeran> or just bump it and wait for the CVEs to roll in
[02:29:47 CEST] <guest4377> so what is the actual maximum usable?
[02:31:08 CEST] <guest4377> how do I find out how much is INT_MAX in my system? or is it a variable somewhere in ffmpeg?
[02:31:23 CEST] <kepstin> the actual maximum usable is such that the (height + 128) * (width + 128) is less than 268435455 on most systems ffmpeg supports
[02:31:50 CEST] <guest4377> oh well
[02:32:02 CEST] <guest4377> that's going to require some maths from me to calculate
[02:32:07 CEST] <guest4377> anyways, thank you
[02:32:32 CEST] <kepstin> which more or less means, the image + overhead of a 64bit per pixel image would fit in 2gb
[04:27:49 CEST] <k_sze> Does anybody know some kind of push-based streaming container/protocol (that FFmpeg happens to support out of the box)?
[04:28:27 CEST] <k_sze> As in, if I want to encode some video and I want to *push* the video stream to a remote computer.
[04:31:29 CEST] <k_sze> Instead of the remote computer pulling (protocols like RTSP/RTP are pull-based, if I understand correctly).
[06:19:09 CEST] <Max-P> k_sze_: There's the udp:// and tcp:// raw protocols
[06:20:11 CEST] <Max-P> And rtmp:// now that I think of it
[06:20:24 CEST] <k_sze_> rtmp can be push-based?
[06:20:32 CEST] <Max-P> Yes
[06:20:59 CEST] <Max-P> I have an nginx server listening to connections, and when I want to stream I ffmpeg to it
[06:21:53 CEST] <Max-P> In this case it's to rebroadcast it, but it could just save it to a file or whatever
[06:21:57 CEST] <k_sze_> Your nginx server is serving a stream or receiving a stream?
[06:22:07 CEST] <Max-P> Both
[06:22:33 CEST] <k_sze_> So you use ffmpeg to push a stream to the nginx server using rtmp.
[06:22:42 CEST] <Max-P> It receives it from me, dump it to a file and redistribute it to anyone connected to it.
[06:22:48 CEST] <k_sze_> interesting
[06:23:03 CEST] <Max-P> Ffmpeg may have one built-in, need to check
[06:23:12 CEST] <k_sze_> rtmp as in this, right? https://en.wikipedia.org/wiki/Real_Time_Messaging_Protocol
[06:24:22 CEST] <Max-P> yes
[06:25:09 CEST] <Max-P> There's that for the client side (ffmpeg): https://trac.ffmpeg.org/wiki/StreamingGuide
[06:25:58 CEST] <Max-P> What do you want to do exactly? The computer that receives it, what does it do with it?
[06:26:34 CEST] <k_sze_> The computer that receives it is supposed to do some real-time machine vision on the stream.
[06:27:34 CEST] <Max-P> And how does that vision program work, you have to push a stream into it?
[06:28:01 CEST] <k_sze_> yes, that's it idea.
[06:29:41 CEST] <Max-P> Is it something that exists (and already listens somehow on a port and expects a specific format), or is it something you are making with the ffmpeg library?
[06:33:31 CEST] <Max-P> Also, just found it, it does support listening on both sides, so it works for ffmpeg to ffmpeg directly for sure. "Point to point streaming" in the wiki page I linked. rtp:// rtsp:// tcp:// udp:// all with the ?listen flag
[06:38:45 CEST] <k_sze_> something I'm going to make.
[06:39:05 CEST] <k_sze_> Which is why I would like to use some existing protocol if there is a suitable one.
[06:39:25 CEST] <k_sze_> nice
[06:41:51 CEST] <Max-P> So yeah, udp/tcp/rtp/rtsp/rtmp are all valid protocols, pick the one that fits the environment the best
[06:43:55 CEST] <k_sze_> There is no relationship between these protocols and container formats, right? So I encode the frame into packets, mux the packets into a container format, and use one of these protocols to stream the container?
[06:45:18 CEST] <Max-P> Yeah I think, I don't know how it works in the ffmpeg library
[07:19:03 CEST] <solrize> icecast?
[07:29:25 CEST] <zumba_addict> is my boss right that when video source has drm, the video has to be completely downloaded first before we can watch it?
[07:30:15 CEST] <Max-P> zumba_addict: no. Otherwise you'd have to download everything entirely on Netflix before you watch it ;)
[07:30:18 CEST] <ln-> not that i know anything, but i strongly doubt that's true for all kinds of videos and drm.
[07:30:33 CEST] <ln-> and indeed Netflix would have been my counter example :)
[07:30:40 CEST] <zumba_addict> ok
[07:30:59 CEST] <zumba_addict> now I need to make a case to prove that what he thinks is wrong
[07:31:11 CEST] <zumba_addict> or what he currently know is wrong
[07:31:37 CEST] <zumba_addict> but i need to present it in a nice way which won't make him feel, I'm stepping on his skills
[07:33:50 CEST] <Max-P> Ask him to find one example of DRM that requires to load the entire file to decode it
[07:34:31 CEST] <Max-P> Because I can't find one actually. Flash, silverlight, DVD, BluRay, ... none of them needs to read the entire thing to open
[07:35:07 CEST] <zumba_addict> k
[07:35:18 CEST] <Max-P> It doesn't make sense for this to even have been a thing at one point
[07:35:48 CEST] <zumba_addict> got it
[07:36:16 CEST] <Max-P> encryption/decryption has always been in a way that you only work on small chunks of data at a time, I don't know of any crypto algorithm that would work on the entire data at once
[07:36:33 CEST] <zumba_addict> do videos with DRM still have moov atom?
[07:36:36 CEST] <aldnavleech> how do you rescale a video into 16:9 aspect ratio, and put a black padding?
[07:37:29 CEST] <Max-P> aldnavleech: Why do you want to do that? Players usually add the black borders themselves
[07:37:30 CEST] <aldnavleech> i still cant seem to get it done
[07:38:09 CEST] <aldnavleech> Ah, I just want to do it during encoding
[07:38:18 CEST] <aldnavleech> especially avconv
[07:38:53 CEST] <Max-P> What's your problem, is it that the video appears stretched? Because adding useless black border usually just use space for no reason and make it annoying to display on different screens
[07:40:01 CEST] <Max-P> It made sense long ago to make VHS and DVDs because the format was fixed, but it doesn't make sense nowadays as any sane video player will auto fit it the best possible in the screen
[07:40:44 CEST] <aldnavleech> this is for web/mobile actually
[07:41:37 CEST] <aldnavleech> just like how youtube does to videos
[07:41:58 CEST] <Max-P> YouTube doesn't add black bars, they even specifically say not to in their guidelines
[07:41:59 CEST] <aldnavleech> i would want to do the same
[07:42:11 CEST] <aldnavleech> ah.
[07:42:14 CEST] <aldnavleech> thanks for the info
[07:42:18 CEST] <Max-P> It just transfers the actual video, and just fit it appropriately in the player
[07:42:32 CEST] <Max-P> And so does any of the popular mobile platforms
[07:43:00 CEST] <aldnavleech> "YouTube uses 16:9 aspect ratio players. If you are uploading a non-16:9 file, it will be processed and displayed correctly as well, with pillar boxes (black bars on the left and right) or letter boxes (black bars at the top and bottom) provided by the player. If you want to fit the player perfectly, encode at these resolutions:"
[07:43:21 CEST] <Max-P> Yes, but it's done on the client side
[07:43:23 CEST] <aldnavleech> https://support.google.com/youtube/answer/1722171?hl=en
[07:44:13 CEST] <Max-P> Yes, that's their recommendation so if you can, you record at that resolution
[07:44:48 CEST] <Max-P> If you don't, then you just give them the file in the correct resolution. You don't add the bars yourself
[07:45:45 CEST] <aldnavleech> but what if i want it in the encoding anyway? :)
[07:46:07 CEST] <Max-P> What you're asking is exactly against their recommendation
[07:46:46 CEST] <aldnavleech> its okay
[07:47:01 CEST] <Max-P> But if you really want to do it, you can use the "pad" video filter: http://superuser.com/questions/26416/how-to-convert-a-169-movie-to-a-43-letterbox-version
[07:47:04 CEST] <aldnavleech> let's ignore the youtube part
[07:48:12 CEST] <Max-P> But keep in mind it's a terrible idea. You don't want black bars in the final video file for any reason. It's wasting space and incredibly inconvenient if you suddently need a different aspect ratio
[07:49:17 CEST] <Max-P> As I said, any sane video players including mobile will play 4:3 videos on their 16:9 or 16:10 screen perfectly, they will add their own bars. It's a trivial calculation, you don't need to think about it
[08:06:14 CEST] <k_sze_> Why does AVFrame.best_effor_timestamp exist?
[08:10:09 CEST] <k_sze_> I mean, what's wrong with AVFrame.pkt_pts?
[08:11:30 CEST] <k_sze_> Is it because pkt_pts can be the same for multiple frames if the frames come from the same packet?
[08:11:59 CEST] <k_sze_> And so ffmpeg would interpolate a proper best_effort_timestamp for each frame?
[08:33:17 CEST] <mtcjayne> When I use ffmpeg my CPU usage climbs to 100%. Is there a way to use my video card for processing instead of my CPU, or is this something implementations of FFmpeg like VLC have to do?
[08:40:59 CEST] <k_sze_> mtcjayne: depends on the codec in question. Some codecs can benefit from hardware acceleration, others can't.
[08:41:43 CEST] <k_sze_> and so it depends on whether your copy of ffmpeg is built with the proper configuration, and whether the video in question is using one of those codecs that can benefit from hardware acceleration.
[08:41:44 CEST] <mtcjayne> I am using h264
[08:42:34 CEST] <k_sze_> h264 can benefit somewhat from hardware acceleration, especially if you have an NVidia video card and your ffmpeg has nvenc enabled.
[08:43:07 CEST] <k_sze_> For certain chroma subsamplings, it can also benefit from Intel QuickSync.
[08:44:48 CEST] <mtcjayne> I am using a Radeon R7 260X, and my build information is here: http://pastebin.com/2BcF5AqM
[08:45:41 CEST] <k_sze_> Radeon R7 should support OpenCL, so you should still be able to benefit from OpenCL for lookahead operations.
[08:46:26 CEST] <k_sze_> (I'm just regurgitating stuff I have read here and there about hardware acceleration in the past months. I'm no expert.)
[09:09:15 CEST] <mtcjayne> Hmm so since I'm rendering at about 50-100 FPS on a concat operation I can't improve that or reduce my CPU usage noticably
[09:15:28 CEST] <k_sze_> If I want to convert from one pixel format to another that is basically just a byte rearrangement (e.g. yuv422p to yuyv422), what flag should I supply to sws_getContext? SWS_BITEXACT?
[11:33:28 CEST] <thecoolguy> Is there a way to make mpeg-dash seekable ?
[12:02:12 CEST] <thecoolguy> anyone ?
[12:14:31 CEST] <jules> Hi, I'm currently playing with the ffmpeg C api. Is there a way to get the current bitrate on packet level?
[12:34:22 CEST] <thecoolguy> is there there a way to have ffmpeg encode to a fileformat that is playable over http while it is transcoding ?
[12:43:04 CEST] <DHE> HLS is probably the best option there, but there's quite a bit of latency introduced.
[12:50:07 CEST] <k_sze_> Hmm, I think I'm still getting *some* memory leak using ffmpeg.
[12:50:31 CEST] <k_sze_> I now make sure to av_free_packet, and that cut down most of leak.
[12:51:45 CEST] <k_sze_> Is there any other place where I can leak if all I'm doing is decode and convert pixel format?
[12:52:56 CEST] <thecoolguy> dhe
[12:53:15 CEST] <thecoolguy> i tried hls with nginx rtmp module but it only seems to support live streaming and not vod
[12:54:01 CEST] <DHE> oh... if the intention is VOD then HLS might be a good choice then. you said "while transcoding" which makes me think of live events only
[12:54:35 CEST] <DHE> it's worth a spin. whether you like it or not is another issue. it has its ups and downs
[13:15:46 CEST] <thecoolguy> DHE yea i know
[13:16:01 CEST] <thecoolguy> the thing is i want the hls to be seeakable
[13:16:19 CEST] <thecoolguy> and it doesn't seem like the rtmp module for nginx supports that
[13:42:19 CEST] <maksimka> ffmpeg keeps freezin on my machine any advice? http://pastebin.com/MqXK0Tpd
[13:55:58 CEST] <thecoolguy> is it possible to get ffmpeg to save to a mp4 chunk ever 10 seconds or something ?
[14:07:15 CEST] <DHE> thecoolguy: -hls_list_size X (where X should be 0 for content that's available start to end)
[14:08:51 CEST] <thecoolguy> I'm using  -ac 6 -strict -2 -c:a aac -c:v libx264 -preset ultrafast -f mp4 segment -segment_times 10,20 -c copy -map 0 "test.mp4
[14:08:56 CEST] <thecoolguy> but it doesn't seem to be spliting the file at all
[14:09:45 CEST] <DHE> I think you want -f segment -segment_format mp4
[14:11:17 CEST] <thecoolguy> like this ? -f segment -segment_times 10,20 mp4
[14:12:12 CEST] <DHE> no
[14:13:54 CEST] <maksimka> any idea what I am doing wrong here? http://pastebin.com/MqXK0Tpd it keeps freezing ..
[14:14:31 CEST] <thecoolguy> exactly as you wrote it ?
[14:27:12 CEST] <thecoolguy> ?
[14:39:19 CEST] <maquefel> guys why i get too fast streaming from mjpeg  file ? ffplay playes normal
[14:39:58 CEST] <maquefel> streaming with ffmpeg -vcodec mjpeg -f mjpeg -i 1.avi -codec mjpeg -r 25 http://localhost:8080/webcam.ffm
[14:41:00 CEST] <maquefel> ffserver settings :
[14:41:01 CEST] <maquefel> https://bpaste.net/show/a22ac5e3c9ac
[15:02:28 CEST] <thecoolguy> How do i fix  Codec for stream 1 does not use global headers but container format requires global headers
[15:02:40 CEST] <thecoolguy> i'm doing  -ac 6 -strict -2 -c:a aac -c:v libx264 -preset ultrafast -map 0 -c copy -f segment -segment_time 5 -reset_timestamps 1 out%02d.mp4
[15:46:46 CEST] <Polochon_street> Hi! I'm decoding songs into an array with this function (http://pastebin.archlinux.fr/1475441), but after that when I access elements in my array, I have randoms (not in every songs) Conditional jump or move depends on uninitialised value(s) with Valgrind
[15:47:01 CEST] <Polochon_street> could my decode function insert these values? Is it fatal?
[15:50:35 CEST] <Polochon_street> maybe the memcpy(index*song->nb_bytes_per_sample + beginning, decoded_frame->extended_data[0], data_size); is invalid, but I believe that if I obtained data_size by av_samples_get_buffer_size, it valid, isn't it?
[16:14:18 CEST] <gonza_> hi people someone can help me with one question i have an error in my code when i try to allocate an AVPicture to buffer with avpicture_layout
[16:14:40 CEST] <gonza_> i have an error in av_image_fill_linesizes+112
[16:32:06 CEST] <thecoolguy> anyone know the reason behing my global header issue ?
[16:44:21 CEST] <thecoolguy> -ac 6 -c:a aac -c:v libx264 -preset ultrafast -map 0 -flags:v +global_header -c copy -g 5 -f segment -segment_time 10 -reset_timestamps 1 mov%02d.mp4
[16:44:31 CEST] <thecoolguy> i tried force it and it still gives me that error
[17:22:42 CEST] <klaxa> thecoolguy: that -c copy looks incorrect
[17:25:23 CEST] <Golfgeo> Hi all
[17:25:46 CEST] <thecoolguy> hm
[17:26:45 CEST] <Golfgeo> I've been playing around with ffmpeg and need to set specific options of my webcam from the commandline command I use to start ffmpeg. It's a h264 stream I'm getting and I need to set h264 specific settings within the cam. Any pointers, tips or a good howto on this?
[17:30:02 CEST] <Golfgeo> Basicly I've got one issue left and that is to change the baseline profile to a high profile so I don't need the scaler to go from a 420 to 422 profile
[17:43:31 CEST] <thecoolguy> klexa any suggestions ?
[17:46:03 CEST] <Golfgeo> Oh just so you know, this is for the "wonderfull" c920 and it's 264 stream... Yea, got it working even to a v4l2 loopback
[17:48:56 CEST] <Golfgeo> it needs about a 3 ghz cpu core/thread to do the convertion. So that's the thing I'm going after with the other profile by not needing any convertion with the scaler
[17:52:51 CEST] <Golfgeo> Anyone?
[17:53:42 CEST] <c_14> What do you mean set options?
[17:53:51 CEST] <c_14> If you want to set options for the camera, use v4l2-ctl
[17:54:18 CEST] <Golfgeo> can they be embedded into the ffmpeg command?
[17:57:35 CEST] <c_14> You can try setting -standard or -input_format
[17:59:00 CEST] <Golfgeo> Am using input_format, but basicly I need to tell v4l2 to set a number of stream specific options in the cam itself
[17:59:29 CEST] <c_14> Then you're probably going to have to use v4l2-ctl
[18:00:07 CEST] <thecoolguy> c_14 I'm using  -ac 6 -c:a aac -c:v libx264 -preset ultrafast -map 0 -c copy -g 5 -f segment -segment_time 10 -reset_timestamps 1 mov%02d.mp4
[18:00:25 CEST] <thecoolguy> and the files anre spliting fine and it is playable on my computer but for some reason it nots playable in ios
[18:01:02 CEST] <c_14> Try adding -pix_fmt yuv420p
[18:01:06 CEST] <Golfgeo> will have a look at the v4l2-ctl options I have, thanks mate
[18:01:50 CEST] <c_14> thecoolguy: or maybe -profile:v main
[18:01:57 CEST] <c_14> https://trac.ffmpeg.org/wiki/Encode/H.264#Compatibility
[18:02:46 CEST] <thecoolguy> I've trid -profile:v baseline -level 3.0 with the same results .
[18:03:04 CEST] <c_14> And the pix_fmt ?
[18:03:10 CEST] <thecoolguy> about to try that now
[18:06:51 CEST] <thecoolguy> yea still not playable. yet when i use certain inputs it is one sec let me get the codec info
[18:11:49 CEST] <thecoolguy> This out worked
[18:11:58 CEST] <thecoolguy> http://i.imgur.com/XtAoItS.png
[18:12:49 CEST] <thecoolguy> while this didn't http://i.imgur.com/fe5f6mR.png
[18:13:38 CEST] <thecoolguy> could it be that there are settings that are to high ?
[18:16:18 CEST] <c_14> hmm, could you provide the ffprobe output of those files? Can't really see anything interesting in that output.
[18:17:14 CEST] <thecoolguy> ok one sec
[18:18:52 CEST] <thecoolguy> any flags or just ffprobe file ?
[18:18:57 CEST] <c_14> just ffprobe file
[18:19:59 CEST] <thecoolguy> Working file http://pastie.org/private/luhkvdlqydwaete6mfkzag
[18:21:04 CEST] <thecoolguy> file broken on ios http://pastie.org/private/j3uhpcx4cawoioyvw2kw4a
[18:23:40 CEST] <c_14> Maybe the audio or subtitle track?
[18:23:47 CEST] <c_14> That or one of the metadata tags
[18:28:08 CEST] <thecoolguy> hm how to a speicify to remove the audio and ubstile chanel using flags ?
[18:28:17 CEST] <thecoolguy> how do i*
[18:29:52 CEST] <c_14> Either explicitly map the video track or use -an -sn
[18:32:24 CEST] <w33t_> howdy! we are trying to compile ffmpeg for ios and we are getting an error saying that librtmp isnt found.
[18:32:38 CEST] <w33t_> we are working off of this: https://github.com/chrisballinger/FFmpeg-iOS
[18:33:02 CEST] <w33t_> it directed us to this room in the build script lol
[18:34:24 CEST] <c_14> Do you need librtmp?
[18:37:01 CEST] <thecoolguy> tried "-an -sn -map_metadata" and still doesn't work. Is there a flag to recalculate duration ?
[18:37:35 CEST] <c_14> hmm?
[18:39:21 CEST] <thecoolguy> Thats the only thing i can think of even after remove the audio, subtitle , and metadata it isn't playing
[18:39:37 CEST] <thecoolguy> the prob is this now http://pastie.org/private/n1jtczeo9agzctnlgglca
[18:40:10 CEST] <thecoolguy> thats pretty much the same as the working example except the duration is wrong
[18:40:29 CEST] <thecoolguy> o wait no it is right
[18:41:12 CEST] <thecoolguy> see any important differences ? they look exactly the same to me
[18:41:34 CEST] <thecoolguy> wel besides bitrate and duration
[18:42:03 CEST] Action: c_14 doesn't notice anything besides resolution and fps
[18:42:09 CEST] <thecoolguy> makes no sense x.x
[19:39:46 CEST] <NapoleonWils0n> hello all
[19:40:09 CEST] <NapoleonWils0n> can ffmpeg use an http proxy ?
[19:40:29 CEST] <NapoleonWils0n> thought i saw something about setting http_proxy somewhere
[19:40:47 CEST] <c_14> There's the httpproxy protocol
[19:41:44 CEST] <NapoleonWils0n> c_14 so can you set http_proxy=something and ffmpeg will use that proxy
[19:42:14 CEST] <DHE> well, you can specify one that way, but c_14 meant there's an input format called "httpproxy" like you have a format for "mp4"
[19:42:25 CEST] <DHE> okay, that's not quite right...
[19:43:23 CEST] <NapoleonWils0n> DHE right so you can get ffmpeg to use an http proxy by setting an env variable for http_proxy is that right
[19:43:27 CEST] <c_14> Not sure how the httpproxy protocol is supposed to be used, but it looks like there is support for the http_proxy environment variable
[19:45:57 CEST] <NapoleonWils0n> any ideas how to set that on linux
[19:46:14 CEST] <NapoleonWils0n> env http_proxy="http://127.0.0.1:6883"
[19:46:42 CEST] <c_14> export http_proxy="http://127.0.0.1:6883"
[19:47:04 CEST] <NapoleonWils0n> right will give it a go
[19:47:37 CEST] <NapoleonWils0n> does ffmpeg pick that up automatically or do i need to pass another argument to ffmpeg
[19:48:48 CEST] <c_14> picks it up automatically
[19:52:14 CEST] <NapoleonWils0n> c_14 i think i love you mate
[19:52:20 CEST] <NapoleonWils0n> just got it working
[19:52:59 CEST] <NapoleonWils0n> actually very cool this works
[19:54:00 CEST] <NapoleonWils0n> just worked out how to record nbc sports with ffmpeg using tinyproxy and adding x-forward header
[20:48:31 CEST] <Guest49918> hi
[20:48:53 CEST] <jwp> i have a question about ffprobe and metadata extraction
[20:49:02 CEST] <jwpppppp> when using ffprobe for metadata extraction, why are properties like "duration", "start_time", "sample_rate" represented as strings when specifying JSON as output?
[20:53:25 CEST] <ChocolateArmpits> eh, What else should they be ?
[20:55:34 CEST] <jwpppppp> ChocolateArmpits: duration and start_time should be a JSON number (float representation) and sample_rate also a JSON number but with integer representation
[20:55:38 CEST] <jwpppppp> maybe I'm missing something?
[20:56:07 CEST] <jwpppppp> e.g. "width" and "height" are JSON numbers (integers) on the JSON output
[21:10:19 CEST] <luc4_> Hello! Is there a list of the accepted params for cpu in configure?
[21:10:30 CEST] <luc4_> I meant accepted values sorry.
[21:39:23 CEST] <claz> is there any way ( using filters or similar ) to add a single frame to the beginning of the video stream?
[21:39:43 CEST] <claz> i have a lot of segments from the same camera that I want to use vidstab on
[21:40:19 CEST] <claz> and i need to use vidstab in 'tripod' mode so it takes an input frame of the video which it uses to offset the entire video
[21:40:37 CEST] <claz> but if i use the first frame of each segment then every segment will be a little different..
[21:41:28 CEST] <ChocolateArmpits> luc4_: most likely not, most systems even when they have matching hardware might have different software running at a single time. There can only be broad generalizations, nothing concrete suggested
[21:42:22 CEST] <pzich> claz: you might be able to use this, but I don't know how well it works when your input is a single frame: https://trac.ffmpeg.org/wiki/Concatenate
[21:43:01 CEST] <claz> thanks, i'm going to try it
[21:44:03 CEST] <luc4_> ChocolateArmpits: I dont understand sorry. You mean I cannot know what to place there for a specific hardware?
[21:44:12 CEST] <pzich> claz: also, make sure to note that there are two sections, "Concatenation of files with same codecs" and "Concatenation of files with different codecs", you may be able to encode your image as a single-frame video of the same codec, or you may have to follow the second part
[21:44:52 CEST] <ChocolateArmpits> luc4_: There is no suggestion list for specific hardware and software setup
[21:45:30 CEST] <luc4_> ChocolateArmpits: if I dont put it it seems to work anyway& should I just not add it then?
[21:45:47 CEST] <ChocolateArmpits> luc4_: What do you mean by "it" ?
[21:46:03 CEST] <luc4_> ChocolateArmpits: - - cpu switch
[21:46:55 CEST] <luc4_> ChocolateArmpits: the proper CFLAGS are sufficient?
[21:48:09 CEST] <ChocolateArmpits> luc4_: What codec are you using ? I'm not finding the 'cpu' switch in ffmpeg's documentation . The cflags is told to only be used for testing.
[21:49:07 CEST] <luc4_> ChocolateArmpits: codec? Sorry& maybe I was not clear. Im building ffmpeg libs for arm. That switch is in the configure script docs.
[21:49:29 CEST] <luc4_> ChocolateArmpits: just run ./configure help and youll get it.
[21:49:39 CEST] <ChocolateArmpits> luc4_: ok, I got confused, I was thinking you were trying to run ffmpeg
[21:49:49 CEST] <luc4_> ChocolateArmpits: no no, building the libs.
[21:50:22 CEST] <luc4_> ChocolateArmpits: just wanted to know cause Ive seen that set in some cases& but not sure what the relationship is to march for instance.
[21:50:32 CEST] <luc4_> Probably more optimization?
[21:51:25 CEST] <llogan> yuo can view fate.ffmpeg.org and look at the various arm builds to see some configure examples. keep in mind that many are configured for fate testing, so for example you wouldn't need --enable-memory-poisoning --enable-avresample
[21:59:45 CEST] <Polochon_street> is avcodec_close(context) enough to prevent memory leaks from allocated codeccontext/codec?
[22:06:22 CEST] <llogan> Polochon_street: might want to ask libav-user mailing list. the mailing list for the FFmpeg libraries
[22:07:55 CEST] <Polochon_street> llogan: don't want to bother a whole mailing for this question, but thanks :)
[22:08:40 CEST] <Polochon_street> +I
[22:09:47 CEST] <llogan> that's what it is there for and it's not nearly as high volume as the others.
[22:10:23 CEST] <Polochon_street> okay, I'll do it then. Thanks!
[22:13:50 CEST] <claz> pzich: that works ! however the problem remains that each first frame of each segment is wrong, any way to remove it without transcoding?
[23:21:26 CEST] <CraziFuzzy> Does anyone know a way to have ffmpeg write to a circular buffer file?  Some way to set a filesize, and it starts writing at the beginning, then when it reaches the end it continues overwriting from the beginning again?
[23:21:43 CEST] <CraziFuzzy> This would be an mpegts stream in the circular buffer.
[23:22:57 CEST] <DHE> there is the "segment" format, or a special case to use "hls" which has some additional baggage (.m3u8 file)
[23:23:31 CEST] <DHE> it writes to multiple files instead, but they concatenate to produce a fine usable video. use -hls_flags delete_segments as well
[23:24:03 CEST] <DHE> oh, segment supports wrapping. even better
[23:34:46 CEST] <CraziFuzzy> I had seen segment, and honestly, never even thought it would support wrapping.  that sounds pretty much exactly what I'm looking for.
[23:37:55 CEST] <CraziFuzzy> so it looks like using a streamsegment format with a wrap limit of 1 might do what i want.
[23:38:51 CEST] <CraziFuzzy> actually, it looks like the segment size is based on duration, not filesize, so it won't really work for my purposes.
[00:00:00 CEST] --- Fri Aug 28 2015


More information about the Ffmpeg-devel-irc mailing list