[Ffmpeg-devel-irc] ffmpeg.log.20140703

[burek]

[00:06] <declan> I have a question relating to the ffmpeg libav API
[00:06] <declan> is this the correct place to ask or is this for CLI only?
[00:07] <c_14> You can ask here, but I don't know anything about the API so you'll have to wait for someone else to answer.
[00:07] <declan> ok
[00:08] <declan> i ave drafted it as i was going to send it to mailing list also
[00:08] <declan> I have a problem which is causing some issues for me. Im using the ffmpeg, libav API within my application.
[00:08] <declan>  I have an input MP4 file with an H264 video stream and an AAC audio stream.  The input video stream is encoded at High profile.
[00:08] <declan> I transcode these two streams at a lower bitrate and then output to an MP4 container.   I also output the H.264 profile to baseline (no B frames).
[00:08] <declan> The  the last 2 frame I send to the decoder are P then B frame. After input is EOF I call the decoder with NULL packet to flush out the remaining frames.  This causes the P frame (2nd last frame) to be returned and marked as complete which I can then encode and mux out, The decoder then returns the last frame (B) to be returned.  However when I encode this one its marked as incomplete and...
[00:08] <declan> ...cant be muxed out.
[00:08] <declan> At this point I then attempt to flush out all delayed frames in the encoder by calling encode with a null picture type to flush the frames out.  However the last frame is marked as incompelte and cannot be written.
[00:08] <declan> Is there anything I can do to get this frame to be output
[00:09] <declan> a rather long question
[00:09] <declan> any help appreciated
[00:09] <c_14> You can't have the last frame be a B-frame.
[00:09] <c_14> The B-frame stands for bi-directional
[00:09] <c_14> It needs a frame before and after for it to be generated correctly.
[00:10] <c_14> s/generated/decoded
[00:10] <LeonBo> I'm having some troubles with -force_key_frames 'expr:gte(t,n_forced*5)'
[00:10] <LeonBo> Every 1 in 4 times I get 'Invalid duration specification for force_key_frames: expr:gte(t'
[00:11] <LeonBo> The input is rtmp and the output also
[00:11] <declan> c_14: so the input file I have has the last frame  as B frame
[00:12] <declan> stream analyzer doesnt indicate any issue with this frame
[00:12] <declan> though maybe thats a problem it has
[00:12] <c_14> declan: you can't decode the last frame. Well let's say shouldn't be able to, I'm not an expert on the topic, but in order to decode a b-frame you need a frame in front and behind.
[00:12] <c_14> Can you view/decode the frame?
[00:13] <declan> yes
[00:21] <declan> c_14: ran it through h264 visa there and I can see the B sample 177 is decoded as frame 176 and sample 176 (P frame) is decoded as frame 177
[00:21] <c_14> It decodes the b-frame before the p-frame?
[00:24] <c_14> LeonBo: try adding a '\' before the ',' ? But seeing as that's how it's written in the manual, not sure what else it could be.
[00:25] <LeonBo> c_14: without the \ it works 3 out of 4 times
[00:25] <LeonBo> Just sometimes it doesn't :(
[00:25] <c_14> Maybe looking at the output will help?
[00:27] <declan> c_14: yea, as u said bi-directional
[00:28] <c_14> declan: If I read your question correctly, the 2nd to last frame was the p-frame and the last was the b-frame, shouldn't it decode the 2nd to last frame before the last one? (Also technically it's bi-predictive, I keep getting the name wrong)
[00:29] <declan> i believe the DTS/PTS stamps tell the player
[00:29] <declan> when to decode and present the frame
[00:30] <declan> though Im no expert, thats how I understand it to work
[00:30] <c_14> Sure, but why would you place a frame you need to decode another frame behind said frame?
[00:30] <c_14> I also tried finding a copy of the h.264 spec to see what it says about the gop and if it can end on a b-frame, but don't really feel like buying a copy from the ITU.
[00:31] <c_14> You might want to wait for someone else, this is a bit too deep for me.
[00:32] <declan> well the encoder may cache a number of frame for optimun compression , quality
[00:32] <declan> hehe
[00:32] <declan> np, thanks
[00:33] <declan> i sent an email to the list, hopefully somone can comment
[03:44] <alexd> when playing back a wmav2 asf encoded in ffmpeg in a silverlight application, i find that seeking is imprecise on the order of 50 ms. Can I change anything during encoding to make seeking more precise?
[03:44] <alexd> hi everyone, btw
[12:53] <alexvf> where are the sources for ffh264vdpau codec?
[12:55] <sacarasc> https://github.com/FFmpeg/FFmpeg/blob/master/libavcodec/vdpau_h264.c
[13:02] <alexvf> thank you sacarasc, i think i need to dig a bit more to know where the h264 poc is set
[13:08] <luc4> Hello! Is winrt an "officially" supported platform for ffmpeg?
[14:43] <dentex> Hell!
[14:43] <dentex> I meant Hello... sorry
[14:47] <dentex> I have compiled ffmpeg for android to use in an app of mine. Now there's a particular device from which I have a logcat showing an error.
[14:47] <dentex> http://pastebin.com/iW2Ya3kZ
[14:49] <dentex> should be an ARMv7 device. This ffmpeg version has been compiled with EXTRA_CFLAGS="-march=armv7-a -mfpu=neon -mfloat-abi=softfp -mvectorize-with-neon-quad"
[14:49] <dentex> Could it be that this device is not NEON enabled, thus the error?
[15:00] <Mavrik> hmm
[15:00] <Mavrik> pretty much everything that's not Tegra 2 on ARMv7/Android has NEON
[15:01] <Mavrik> and you get a "invalid instruction" crash
[15:01] <Mavrik> what version is your ffmpeg ?
[15:02] <Mavrik> dentex, ah... look at that
[15:02] <Mavrik> you are getting ILLOPC - illegal opcode
[15:02] <Mavrik> dentex, yep, Lenovo K1 has Tegra 2, so no NEON support
[15:02] <dentex> Hello. FFmpeg version's is 2.1.3
[15:03] <dentex> So binaries compiled with NEON support crash on device without NEON?
[15:04] <Mavrik> yes of course
[15:04] <Mavrik> since the CPU doesn't understand NEON instructions.
[15:04] <dentex> ... :O
[15:04] <dentex> great, thanks. At least I know what to try next
[15:05] <dentex> I'll compile a test version with EXTRA_CFLAGS="-march=armv7-a --disable-neon"
[15:05] <dentex> ...ok
[15:05] <dentex>  --disable-neon NOT in EXTRA_CFLAGS...
[15:10] <Mavrik> dentex, -march=armv7-a -mfloat-abi=softfp -mfpu=vfpv3-d16 -I$PREFIX/include -DANDROID
[15:10] <Mavrik> --disable-neon is in ./configure not in EXTRA_CFLAGS.
[15:11] <dentex> Mavrik, thank you very much
[15:18] <dentex> Mavrik, re-compiling right now with your switches... fingers crossed!
[15:52] <alexd457> hi everyone
[15:52] <alexd457> when playing back a wmav2 asf encoded in ffmpeg in my silverlight application, i find that seeking is imprecise on the order of 50 ms. Can I change anything during encoding to make seeking more precise?
[16:33] <Peter_Occ> How do I capture jpegs from a video camera in a local network?
[16:34] <sacarasc> alexd457: It's 0.05 seconds out when you seek?
[17:24] <Maqs> could anybody point me in the right direction on how to convert a H.264 NAL stream (starting with SPS/PPS) to a live mp4 stream for html5/flash in c?
[17:35] <Peter_Occ> I have figured out how to capture video over lan, if I use mpeg4 encoding, it works but when I change ti to h264, it seems to capture the video OK but when I look at it, I just get a black screen. Do you see anything wrong with this command line?  ffmpeg -f mjpeg -r 8 -i http://192.168.0.7/video/mjpg.cgi -vcodec h264 -b 1000000 -r 8 /var/www/html/mcamip/video_file2.avi
[17:39] <alexd457> sacarasc: Sometimes the seek is perfect, sometimes it's off. It basically looks like the position is being rounded. 50ms is a guess. could be 100ms. could vary. hard to say exactly
[17:39] <Mikk36> Hey
[17:39] <Mikk36> Trying to start ffserver, but getting an error about LIBAVFORMAT_53 not being defined
[17:40] <Mikk36> ffserver: relocation error: ffserver: symbol ffm_read_write_index, version LIBAVFORMAT_53 not defined in file libavformat.so.53 with link time reference
[17:40] <Mikk36> am i missing something?
[20:24] <gcl5cp> Why would someone in their right mind would use LIBAV instead of FFMPEG?
[20:24] <sacarasc> You should ask that in #ubuntu or #debian on OFTC.
[20:25] <gcl5cp> i asked without any answer
[20:25] <Hello71> !g why use libav
[20:26] <gcl5cp>  i asked without an answer
[23:57] <jtakacs> hi!
[23:58] <jtakacs> is it possible to use ffmep for rtp streaming, with padding mpegts packets to make it use a constant bandwidth?
[23:58] <jtakacs> *ffmpeg
[00:00] --- Fri Jul  4 2014