[Ffmpeg-devel-irc] ffmpeg.log.20161128

Tue Nov 29 03:05:01 EET 2016

[01:10:24 CET] <hiihiii> what's the ffmpeg command to merge/join two vids which are online? I'm using "concat -i "url_1" -i "url_2" -codec copy output.mp4
[01:22:51 CET] <Mysoft> ok, thanks for all the insights, i finally got my recorder working as best as i can
[01:24:11 CET] <Mysoft> (windows XP / core2quad 2.5ghz / 4gb / 1mbps upload, capturing from the game with my own "injected module", and pipying the i420p+pcm to ffmpeg trough an .avi
[01:24:31 CET] <Mysoft> and so on... tuning the best i could for quality recording of paladins...
[01:24:39 CET] <Mysoft> result https://twitch.tv/mysoft
[01:24:49 CET] <Mysoft> so thank you for all the fish :D
[02:05:28 CET] <FishPencil> Does x264's CRF value actually calculate a "quality" target? As in, if I had two inputs of the same media with only different bitrates, would x264 compress them to the same size, if the same crf value is used?
[02:06:56 CET] <FishPencil> Or is the crf value more of a percentage of the original bitrate?
[02:23:16 CET] <DHE> FishPencil: broadly speaking busier frames will have the qp value raised. the crf setting specifies a sort of "nominal" qp value (that is oversimplifying it a lot)
[02:34:47 CET] <Hello71> how do I make showspectrum go faster?
[02:40:57 CET] <FishPencil> DHE: With identical inputs only varying with bitrate, will the same CRF value produce the same output, or does the input bitrate matter
[02:41:43 CET] <DHE> crf is intended as an alternative to bitrate selection mode
[02:42:41 CET] <DHE> though you can use a sort of constrained mode
[02:43:48 CET] <FishPencil> DHE: So that would imply a "yes" to my question?
[02:46:09 CET] <DHE> I don't know. I've never tried that. I'm saying what I know
[02:47:53 CET] <DHE> that such a mode exists
[02:48:35 CET] <FishPencil> DHE: I'm asking because I'm seeing oddly different file sizes with different streams from the same bluray. I would assume they should all be close to the same, assuming their motion is somewhat similar, but some are 3GB+, and other <1GB
[04:54:25 CET] <mozzarella> guys
[04:54:30 CET] <mozzarella> if I install gstreamer-ffmpeg, will it play all formats ffmpeg can play?
[05:12:09 CET] <tomed> do you actually gain anything by using  slow x264/x265 setting instead of fast
[05:25:17 CET] <thebombzen> tomed: yea. depends on the setting. if you are using CRF then the quality will be the same but slow will have a bitrate improvement.
[05:25:35 CET] <thebombzen> if you're using a VBR mode like -b:v 2000k, then slow will have better quality than fast.
[05:25:45 CET] <tomed> thebombzen  are you sure even if i use  ultrafast setting?
[05:26:22 CET] <thebombzen> tomed: ultrafast is different cause it forces a profile change. but yea, -crf:v 18 will have the same quality at all presets.
[05:26:48 CET] <thebombzen> but if you use ultrafast, the bitrate will be absurdly high and if you use slow, the bitrate will be more reasonable.
[05:26:58 CET] <tomed> so  crf20 on ultrafast and placebo  will same same quality output?
[05:27:04 CET] <furq> no
[05:27:19 CET] <furq> the same crf will look better with slower presets
[05:27:30 CET] <thebombzen> furq: I don't understand?
[05:27:31 CET] <tomed> it shouldn't though
[05:27:38 CET] <thebombzen> afaik that's because ultrafast forces the constrained baseline profile. but everything else should be identical?
[05:28:10 CET] <furq> no, you will often find veryslow results in a bigger file than slow
[05:28:47 CET] <thebombzen> I also don't understand why that will occur?
[05:28:50 CET] <thebombzen> that sounds like a bug.
[05:29:00 CET] <tomed> furq  if that is true, then what is the point of "very slow"
[05:29:14 CET] <furq> like i said, it increases quality at the same crf
[05:29:29 CET] <thebombzen> furq CRF stands for constant rate factor. it's a constant quality.
[05:30:02 CET] <furq> http://forum.doom9.org/showthread.php?t=157779
[05:30:42 CET] <thebombzen> that thread is six years old
[05:31:40 CET] <furq> http://forum.doom9.org/showthread.php?p=1567146#post1567146
[05:34:04 CET] <thebombzen> well that directly contradicts what's on the ffmpeg wiki last time I read it
[05:34:18 CET] <furq> a lot of things directly contradict the ffmpeg wiki
[05:35:00 CET] <thebombzen> well it says on https://trac.ffmpeg.org/wiki/Encode/H.264 that you should choose a CRF value first, and then a preset.
[05:35:04 CET] <thebombzen> if that's wrong, someone should change it.
[06:33:24 CET] <kepstin> I've found that changing the preset while keeping the crf the same usually seems to change the resulting bitrate more than the quality.
[06:34:12 CET] <kepstin> but then i'm mostly comparing 'veryslow' to 'veryfast', don't use the in-between modes much :/
[09:30:04 CET] <Filarius> hi, i'm looking for RGB to YUV formula used by FFMpeg in RGB24 to YUP420P conversion, i'm not so good in diggining source code, just found it must be somethere around swscontext and sws_scale
[09:38:35 CET] <JEEB> the formula itself is generic, you should look at something like BT.709 or BT.2020 or such, as not YCbCr is the same
[09:54:57 CET] <SouLShocK> is it possible to use timestamp from input stream, to name files with the strftime using the timestamp from the input stream when writing to -segment format? i.e. name files output_2016_11_27_22_14_00.ts when input stream is at 22:14:00, instead of using local clock on the machine running ffmpeg
[10:52:59 CET] <techtopia> is there someway i can scan a video for dupe frames eith ffmpeg
[11:13:38 CET] <durandal_1707> techtopia: you can use framemd5 muxer or any hash muxer to help search for dupe frames
[13:48:21 CET] <barhom> Hey, Im using libzvbi and "-txt_page" to overlay some dvb subtitles into the video, but I dont know how to find out the correct teletext (subtitle) page number.
[13:48:26 CET] <barhom> Is there a way to list all available pages?
[14:13:30 CET] <Mathijs> I want to extract audio from a video file and convert it to wav. Is it possible to obtain a wave-file (not raw, so with header) without writing to a file with the .wav extension?
[14:13:31 CET] <Mathijs> I use a batch-processing tool that provides me with a temporary file (which I want to write the audio to) but that tempfile does not end in .wav and ffmpeg seems to write raw samples to it.
[14:14:15 CET] <furq> Mathijs: -f wav
[14:14:51 CET] <Mathijs> That's easy. I googled and couldn't find it...
[15:08:25 CET] <superguest> https://ffmpeg.org/ffmpeg-utils.html#Date
[15:08:56 CET] <superguest> [(YYYY-MM-DD|YYYYMMDD)[T|t| ]]((HH:MM:SS[.m...]]])|(HHMMSS[.m...]]]))[Z]
[15:09:48 CET] <superguest> there's an extra ']' in both (HH:MM:SS[.m...]]]) and (HHMMSS[.m...]]])
[15:10:43 CET] <superguest> in case someone cares
[15:11:00 CET] <JEEB> the docs are in the repo, if you found a bug you can write a patch and have it fixed
[15:11:04 CET] <JEEB> vOv
[15:27:05 CET] <Guest72179> good morning
[16:00:21 CET] <MustBeLucky2> Has anyone experienced and fixed an ASLA buffer xrun problem before? I believe this is making my audio drop.
[16:04:49 CET] <MustBeLucky2> i see a post that says to increase the ASLA_MAX_BUFFER_SIZE but they say its located in libavdevice/asla-audio.h but this file does not exist
[16:05:22 CET] <MustBeLucky2> also, i decreased the frame rate via -r , but the log shows that it starts at 75 FPS and slowly drops down.
[16:07:46 CET] <MustBeLucky2> my command http://pastebin.com/45mkeywS
[16:24:14 CET] <debianuser> MustBeLucky2: It may be not because of alsa, but because of video encoding or otherwise high cpu usage, or slow network maybe. So ffmpeg can't read alsa stream in time, alsa just exposes that to you. Try recording just alsa to a file, e.g. `ffmpeg -f alsa -ac 1 -i hw:1,0 -acodec aac -strict -2 somefile.aac` - do you get xrun then?
[16:26:46 CET] <MustBeLucky2> thank you debian user :) i will run this and see
[16:31:46 CET] <durandal_1707> MustBeLucky2: that define is located in alsa.h
[16:33:19 CET] <MustBeLucky2> Thank you durandal_1707. debianuser , thanks for this. Recording audio only to the file does not give me a xrun. I am going to try audio and video to file now. I think I may have an issue with crtmpserver
[16:35:13 CET] <MustBeLucky2> I wish it would just reconnect when it hits an xrun. it stinks that it drops my audio all together. im sure this is for a reason though
[16:38:05 CET] <MustBeLucky2> I get an xrun when recording video and audio to file.
[16:38:05 CET] <debianuser> "xrun" is just an alsa warning (soundcard/driver writes data faster than ffmpeg can read it), but ffmpeg should automatically recover from it, skip some frames and continue recording. I mean it's not a good thing, but it's not a fatal error, audio stream should not be dropped because of that.
[16:39:00 CET] <MustBeLucky2> strange, something else must becausing the audio to drop then, but the audio drops right around the time i hit the xrun
[16:39:39 CET] <debianuser> check `top` while ffmpeg video+audio is running, maybe you have 100% CPU usage and just don't have enough resources to encode both video+audio at the same time?
[16:41:18 CET] <debianuser> For a test you can try a smaller resolution, e.g. 320x240, it's not too useful for recording, but should use less cpu, check if your audio gets lost after that.
[16:42:31 CET] <debianuser> MustBeLucky2: by the way, you said you get xrun when recording video and audio to file. Now try playing that file you recorded - does the audio gets lost there too?
[16:44:19 CET] <MustBeLucky2> ok perfect thank you. Im surprised im not pulling more CPU on my Pi but it says im only at like 30% tops
[16:44:37 CET] <MustBeLucky2> going to try smaller resolution .... how about bitrate?
[16:44:42 CET] <MustBeLucky2> should i decrease that?
[16:53:38 CET] <MustBeLucky2> so strange. It seemed pretty stable last night, now I cant get it to stay up for a few seconds
[16:57:51 CET] <MustBeLucky2> durandal_1707, do you know if increasing that buffer changes my audio offset?
[16:59:03 CET] <durandal_1707> no, never tried recording on current laptop
[17:10:22 CET] <MustBeLucky2> can ALSA buffer xrun be cause by a crappy and noisy mic?
[17:11:12 CET] <debianuser> no
[17:11:25 CET] <debianuser> and if it could - you'd get it even when you record just audio and save it to a file
[17:11:36 CET] <MustBeLucky2> ah
[17:12:29 CET] <MustBeLucky2> well adding -thread_queue_size 1024 to video and audio input has kept the audio from dropping competely, but it bounces in and out
[17:13:36 CET] <MustBeLucky2> nevermind it finally dropped
[17:13:38 CET] <MustBeLucky2> :(
[17:13:40 CET] <MustBeLucky2> shoot
[17:22:12 CET] <debianuser> is the audio dropped if you save it to a file?
[17:23:11 CET] <alexvf> hi, anyoone knows how can i totally flush a video decoder? avcodec_flush_buffers seems not enough
[17:24:01 CET] <BtbN> recreate it
[17:24:29 CET] <alexvf> is there no other way?
[17:25:02 CET] <alexvf> i would like to reuse the same
[17:26:59 CET] <alexvf> to recreate the AVCodecContext, it seems i have to call avformat_open_input again, which is slow for my needs
[18:06:42 CET] <fred1807> can mp4 be VP9 ?
[19:11:28 CET] <t4nk095> hey there. i got a quick question regarding ffmpeg and subtitles. I need to create a prores master with burned in subtitles. is there any way i can tell ffmepg to output the same video&audio container&codec which is the input? i mean a pass-through but burn in the  subtitles at the same time. also how do i style and position the subtitles? any help is much appreciated :) greetings from munich
[19:12:02 CET] <BtbN> you can't burn in without re-encoding
[19:14:03 CET] <kline> hi, recently ive been having some issues with recording - specifically ffmpeg seems to stop being able to record after the second frame. the last warning is that there may be an issue with my swscaler settings, but i ask it to actually not scale (as far as i can see), so im not sure what to investigate from here. the full log is at https://paste.debian.net/899533/
[19:15:58 CET] <BtbN> something sent an SIGINT, so it exited.
[19:16:03 CET] <kline> you can see on line 38 that something very suspect is going on, but im not sure how to interpret it myself or go about investigating it. this line used to work to record decent mkvs, but im not sure what changed some time in the past month or so in the end
[19:16:19 CET] <kline> and yes, i send sigint after running it for a minute or so to generate the log
[19:16:29 CET] <kline> (via ctrl+c)
[19:16:56 CET] <kline> i can hit Q to stop it as well if you'd rather i do that for logs
[19:18:07 CET] <roxlu> hey! I'm using ffmpeg in my C++ app. When I try to decode a .mp4 file (that was created using Quicktime on Mac), I get lots and lots of errors like: https://gist.github.com/roxlu/ae6e2e4fb06992da91bc8ac36c9ddd6f
[19:18:08 CET] <thebombzen> kline: try using -c copy -f null -. see if it's an encoder issue
[19:19:05 CET] <kline> cool. in the meantime i terminal recorded to show what it looks like, but i dont think its that much more interesting: https://asciinema.org/a/3ouawqojmni2opwyjfdg8b8uk
[19:20:39 CET] <thebombzen> kline: it's also possible it's a problem with your video drivers. what card do you have and what driver are you using?
[19:21:23 CET] <thebombzen> I was having trouble running x11grab with nouveau, just as example. I had to switch to the proprietary nVidia drivers to get it to work nicely.
[19:22:48 CET] <kline> thebombzen, sorry, im not a big user of ffmpeg, can you clarify what that would look like in the context of the original line? as for driver, i have `Intel Corporation Haswell-ULT Integrated Graphics Controller (rev 0b)` as my "card", and `i965-va-driver` for amd, version 1.4.1-2
[19:23:11 CET] <kline> basically, its a fairly standard debian on intel thinkpad x240
[19:23:54 CET] <thebombzen> I mean ffmpeg -f x11grab -video_size 1366x768 -framerate 60 -i :0.0 -c copy -f null -
[19:24:33 CET] <thebombzen> the -c copy says "don't reencode." the -f null is the "null format" which discards all output and doesn't write it to any file.
[19:25:01 CET] <kline> https://paste.debian.net/899535/
[19:26:37 CET] <thebombzen> that sounds like a bug, but it doesn't occur in my FFmpeg build and I build it every night
[19:26:49 CET] <kline> frustratingly, this used to work, but i use ffmpeg infrequently enough that i dont know what has since changed :(
[19:27:14 CET] <thebombzen> try a minimal build from git master and see what happens
[19:27:56 CET] <thebombzen> it's nice that Debian has switched off libav
[19:31:26 CET] <kline> oh, hm, i think that might have been the breaking point. did this happen after the 4th october?
[19:32:05 CET] <kline> i found a video i captured from that date (including build flags i think) but i'm still picking apart the frames to see what version it was etc
[19:34:37 CET] <t4nk095> I know I have to re-encode, I just want the output to be the same codec as the input. prores in -> burn in subtitles -> prores out. can ffmpeg read the detail of the input file and use that to create the output?
[19:34:48 CET] <kline> argh, the version numbers dont appear in any frame, too much is flushed to the console at once and by the next frame theyve fallen off the top
[20:02:45 CET] <c_14> t4nk095: no, you'll have to script that (or use the libraries)
[20:04:16 CET] <t4nk095> I suppose copy stream wont do the job? :D
[20:04:30 CET] <furq> not if you want to burn in subtitles
[20:19:31 CET] <t4nk095> any ideas why this wont work? "ffmpeg -i input.mov -vcodec prores-ks profile: hq -acodec copy -vf subtitles=sub.srt output.mov"
[20:23:19 CET] <c_14> profile: hq isn't an option
[20:23:29 CET] <c_14> did you mean -profile hq ?
[20:23:31 CET] <c_14> maybe
[20:29:37 CET] <t4nk095> I want to encode a ProRes 422 (HQ). But "-profile hq gives me: "Please use -profile:a or -profile:v, -profile is ambiguous"
[20:30:16 CET] <durandal_1707> so add :v
[20:32:47 CET] <t4nk095> which returns "Unknown encoder 'prores-ks'" :D
[20:33:17 CET] <c_14> prores_ks
[20:36:49 CET] <t4nk095> awesome!
[20:37:49 CET] <t4nk095> is there a comparison between prores_ks and_aw somewhere?
[20:42:12 CET] <durandal_1707> ks have finertia control versus quality iirc
[22:47:19 CET] <Mista-D> get constant errors in MPEG2 VIDEO and xerror is not exiting. IF AC3 error is encountered, then xerror works. Please advise if there's any method to increase "xerror" sesitivity to video errors? `ffmpeg -i bad.ts -an -f null -strict very  -xerror -" = no errors detected by xerror due to "-an" switch...
[23:03:18 CET] <Mista-D> "-err_detect +explode " has no effect, exit status code is 0 and xerror is not exiting...
[23:15:52 CET] <Hello71> how do I make showspectrum move faster?
[23:31:19 CET] <Guest94097> can someone tell me what the command is for ffmpeg to taking a mpts and output a spts?
[23:42:22 CET] <lvella> when encoding MPEG-2, what is the smallest unit contained in a AVPacket?
[23:42:28 CET] <lvella> is it a picture?
[23:43:12 CET] <lvella> or a picture may be broken between AVPackets ?
[23:43:25 CET] <DHE> There should be a 1-to-1 correlation between AVPackets and AVFrames (though not necessarily in order)
[23:44:16 CET] <lvella> in interlaced mode, an AVFrame contains one interlaced field?
[23:46:38 CET] <DHE> I believe 1 AVFrame contains both fields
[23:50:49 CET] <JEEB> DHE: that depends on the format
[23:51:12 CET] <JEEB> although if we're talking about mpeg-2 video encoding most likely what you say is true
[23:51:31 CET] <JEEB> HEVC decoder for example outputs fields separately if they are coded as separate pictures (which they mostly are)
[23:52:04 CET] <kerio> ew interlaced video
[23:52:43 CET] <lvella> kerio: not my fault, I didn't write the norms
[23:53:49 CET] <DHE> Keridos: over-the-air broadcasts are almost universally 1080i. frequently in mpeg2 as well...
[23:55:24 CET] <lvella> so, in interlaced mode both fields are inside a single AVPacket?
[23:56:05 CET] <lvella> that would mean fields can't be reordered for better compression, right?
[23:57:03 CET] <lvella> I mean, inside the GOP
[00:00:00 CET] --- Tue Nov 29 2016