[Ffmpeg-devel-irc] ffmpeg.log.20151111

[burek]

[01:26:51 CET] <fred1807> guys how do you write bash scripts for ffmpeg, and pass filenames with spaces, wheh filename derivates from a variables?   "$i"  does not work  when filename has spaces, I need somehow to place the filenames inside a single quote, but thats seems impossible in bash
[01:57:43 CET] <jasom> "$i" should work fine
[07:18:53 CET] <Kota> Hi. I'm using Windows 8.1. When I try to use FFMPEG to convert to a webm, I'm having trouble embedding subtitles. My subtitled input file is an mkv with soft subs (ass/ssa), using -vf subtitle=file.mkv. It states that it's unable to parse option value "file.mkv" as image size.
[07:18:59 CET] <Kota> http://pastebin.com/aqa9huYm my error and my command
[07:30:51 CET] <klaxa> looks like escaping hell
[07:38:58 CET] <Kota> (klaxa) I've heard that it needs escape characters so I tried them. Didn't seem to help
[09:28:42 CET] <flux> c_14, so do you (or someone else) have an idea how long the fate suite takes to run on a i7-2600K CPU @ 3.40GHz? I imagine it's hours, but is it days?-)
[09:30:48 CET] <flux> I guess I'll try to be smart and pick only the relevantish tests :-)
[10:27:39 CET] <anddam> OT question, what's a proper CLI tool to set m4a file metadata?
[10:28:09 CET] <anddam> like id3tool or id3v2 for mp3 files
[10:32:54 CET] <anddam> answer to self: AtomicParsley
[10:32:58 CET] <anddam> sorry for the noise
[11:43:11 CET] <abd5932> Hi, I mix my video and sound like this: ffmpeg -i video.mp4 -i sound.mp3 -c:a aac -strict -2  -shortest video_sound.mp4 . Is there a way to make the sound "fade away" on the last second? Thank you!
[11:44:51 CET] <furq> abd5932: https://ffmpeg.org/ffmpeg-filters.html#afade-1
[11:45:06 CET] <abd5932> furq: thank you!
[11:49:54 CET] <JamJams> Hi! Question for the libavcodec people, is it possible to avformat_open_input a file that's still being created and process it but avoid any EOF errors?
[12:16:49 CET] <JEEB> JamJams: 1) make sure it's transcoding faster than realtime 2) make sure you have let it do at least one GOP complete or so 3) use a format that you can that doesn't require a global index to function
[12:17:01 CET] <JEEB> 4) use whatever the API equivalent of -re is
[12:19:55 CET] <JamJams> JEEB it functions fine atm but I wrote in a dirty method of while (pFormatCtx->pb->pos+1000000) > currentfilesize sleep(10); to avoid hitting the EOF of the input.
[12:20:04 CET] <JamJams> I'm just wondering of if there's a better more native way of doing it
[12:20:07 CET] <JamJams> It seems like there should be
[12:25:03 CET] <DHE> it varies by OS, but you can register for notifications on file size changes/writes. linux uses inotify for example...
[12:25:13 CET] <DHE> but historically the sleep(1) method is what most people did before that
[12:27:15 CET] <JamJams> Yeah, that's not really my issue though my issue is that avformat_open_input can't handle a file that's still been written to
[12:27:58 CET] <JamJams> Perhaps if I made a custom IO-Context that read from the file indefinitely
[18:29:32 CET] <Kota> Hi. I'm using Windows 8.1. When I try to use FFMPEG to convert to a webm, I'm having trouble embedding subtitles. My subtitled input file is an mkv with soft subs (ass/ssa), using -vf subtitle=file.mkv. It states that it's unable to parse option value "file.mkv" as image size.
[18:29:35 CET] <Kota> http://pastebin.com/aqa9huYm my error and my command
[18:42:36 CET] <waressearcher2> Kota: hallo
[18:43:17 CET] <Kota> 'lo
[18:49:25 CET] <waressearcher2> Kota: wie gehts ?
[19:01:51 CET] <Kota> now I'm getting an error that doesn't exist
[19:02:20 CET] <Kota> I've extracted the .ass and am using that as input to a complex filter, getting cannot find a matching stream for unlabeled input pad 0 on filter parsed_subtitles_1
[19:03:45 CET] <c_14> Kota: afaik you can't put a semicolon in a -vf
[19:03:51 CET] <c_14> You have to use a filter_complex
[19:04:19 CET] <c_14> Also, your escaping is probably wrong.
[19:04:32 CET] <Kota> Here, let me paste my new command and error
[19:05:05 CET] <Kota> http://pastebin.com/pMu4gb4v
[19:06:33 CET] <c_14> replace the ';' with a ','
[19:06:48 CET] <c_14> You could probably do the same with your original command
[19:07:01 CET] <Kota> oh nice. I'll try that in a moment
[19:07:14 CET] <Kota> To be honest I have little understanding of half the flags and syntax
[19:07:22 CET] <Kota> I just make it work based on what I've seen before
[19:07:58 CET] <furq> you will probably have an easier time all round if you put ffmpeg on your PATH so you can run it from the directory containing your files and pass relative paths
[19:08:39 CET] <furq> (or run ffmpeg with an absolute path but that's a chore)
[19:11:41 CET] <orbea> What does this aac element type mismatch spam mean in this output using mpv 0.13.0 and ffmpeg 2.8.1 on slackware 64 current? http://dpaste.com/1PYX95W
[19:12:06 CET] <orbea> just freshly recompiled ffmpeg and mpv as well
[19:19:40 CET] <Kota> hmm
[19:19:46 CET] <Kota> It didn't put in the audio from the source file
[19:45:39 CET] <c_14> Kota: try mapping it explicitly
[19:50:07 CET] <Kota> oh derp, webm doesn't support the original stream
[19:50:11 CET] <Kota> had to convert to vorbis
[21:45:25 CET] <SpeakerToMeat> Hello all.
[21:46:38 CET] <SpeakerToMeat> Question, what would be the "right" (better) way with ffmpeg to recode a 25fps material to 23.976? I'm thinking the diff might still merit using pts to reinterpret as 23.976, rather than using the fps vf, or -r. But then how do I match audio?
[21:47:40 CET] <IntelRNG> YOu'd have to change the audio speed. That changes the tones and everything. I have not had a look at it but I guess it is doable.
[21:48:01 CET] <furq> SpeakerToMeat: https://ffmpeg.org/pipermail/ffmpeg-user/2013-February/013406.html
[21:49:02 CET] <furq> it depends on whether the audio was pitch corrected
[21:50:18 CET] <SpeakerToMeat> So yes, setpts is the right way
[21:50:30 CET] <furq> yeah -r will drop frames
[21:50:56 CET] <SpeakerToMeat> I need to read abotu setpts, to see what things like (PTS-STARTPTS) mean
[21:54:40 CET] <SpeakerToMeat> furq: I'm guessing only drop frames is supported and not some sort of optical flow
[21:58:27 CET] <furq> i'm not aware of any frame blending filter
[21:58:37 CET] <furq> that's not what you want for pal -> film anyway
[22:00:46 CET] <SpeakerToMeat> hmm yeah
[22:02:16 CET] <furq> depending on what format they're in, i think mkvmerge can change the video fps without re-encoding
[22:02:22 CET] <furq> you'd still need to process the audio though
[22:03:30 CET] <furq> i've been meaning to do this for a few things but i can't figure out whether the audio has been pitch corrected or not without obtaining a 23.97fps source
[22:03:36 CET] <furq> at which point i might as well just use that instead
[22:04:33 CET] <ChocolateArmpits> "i think mkvmerge can change the video fps without re-encoding" -- That highly depends on the format used
[22:04:43 CET] <furq> it sure does
[22:05:15 CET] <SpeakerToMeat> furq: if I do the audio resampling trick to avoid modification: "-filter:a asetpts='PTS-STARTPTS',atempo=0.959040959" will this work with an "-acodec copy"?
[22:05:22 CET] <furq> no you need to re-encode
[22:05:35 CET] <SpeakerToMeat> furq: No, I need to reencode anyhow, as I'm moving it to prores hq
[22:05:40 CET] <SpeakerToMeat> Ok I'll reencode then
[22:06:36 CET] <furq> i assume this is originally a film source which has been converted to pal
[22:06:41 CET] <furq> otherwise you do want to drop frames
[22:07:35 CET] <furq> s/do/might/
[22:08:16 CET] <SpeakerToMeat> Do you think you can explain this to me? setpts='25025/24000*(PTS-STARTPTS)
[22:08:32 CET] <SpeakerToMeat> I get pts is changing the time (place) value of each frame. right?
[22:09:10 CET] <SpeakerToMeat> I?m guessing PTS-STARTPTS is used in case timecode doesn't start at 0. I guess...
[22:09:11 CET] <ChocolateArmpits> timestamps to be exact
[22:09:36 CET] <furq> that sounds right to me
[22:09:39 CET] <SpeakerToMeat> But why did the person obstensibly doign 25 to 23.976 as I want, chose 25025/24000 as the multiplier?
[22:11:22 CET] <SpeakerToMeat> Instead of, lets say. 25000/23976 for example?
[22:11:52 CET] <ChocolateArmpits> the correct timebase is 24000/1001
[22:11:59 CET] <ChocolateArmpits> for 23.967...
[22:13:03 CET] <SpeakerToMeat> ChocolateArmpits: And that's used as the -r rate in the output...
[22:14:39 CET] <furq> 25025 is 25*1001
[22:15:29 CET] <SpeakerToMeat> Interesting. Ok, and then he (I) uses 24000 as the dividend. I think I get it.
[22:17:22 CET] <SpeakerToMeat> Thank you very much for the help
[22:18:58 CET] <SpeakerToMeat> Little question, I'm transcoding to prores HQ on quicktime wrapper.... and it doesn't accept pcm 16 bit (little endian or big endian), says there's no known tag for the codec in the container... but mov should be able to use 16 bit pcm (it workes with 24bit). Am I crazy or something?
[22:22:57 CET] <SpeakerToMeat> And, give I'm reinterpreting (pts), do I need to remove the timecode stream from the final? (copied from the original).
[22:23:15 CET] <SpeakerToMeat> Or should the timecode stream on the final be not a copy from the original but a new one generated by ffmpeg
[22:27:25 CET] <ChocolateArmpits> SpeakerToMeat: the timecode track has no relation to timestamps. You can use the -timecode option to set it
[22:27:32 CET] <ChocolateArmpits> I think
[22:28:21 CET] <SpeakerToMeat> Interesting, so there wouldn't be a timecode for each frame in the track?
[22:28:45 CET] <SpeakerToMeat> Ah it is the start timecode I guess
[22:28:54 CET] <SpeakerToMeat> Which wouldn't be affected
[22:32:20 CET] <ChocolateArmpits> it might be a case with MPEG TS but generally timecode is not related to timestamps
[22:34:15 CET] Action: SpeakerToMeat nods
[22:34:19 CET] <SpeakerToMeat> Ok, thanks again
[00:00:00 CET] --- Thu Nov 12 2015