[Ffmpeg-devel-irc] ffmpeg.log.20190501

burek burek021 at gmail.com
Thu May 2 03:05:02 EEST 2019


[00:18:35 CEST] <pink_mist> Montana: so you're spamming channels because you're ... what? hired by AMD to advertise their superiority? either way, please never do that again.
[00:28:36 CEST] <de-facto> How do i put an PNG image as video canvas (background) for overlays? Right now i choose black like -i "color=r=60:c=black:s=1600x1200:d=30" but i want the same with a PNG image instead of black...
[00:38:27 CEST] <another> & & &
[00:43:58 CEST] <de-facto> ...simply -loop 1 -i "background.png"?
[00:45:43 CEST] <another> probably
[01:26:01 CEST] <Classsic> Hi everybody
[01:33:03 CEST] <Classsic> here a  little question, I receive rtsp stream and have to write audio and video in two different stdout, but each stream have different timestamp
[03:56:17 CEST] <Montana> https://mp4repair.org  can  playback a mp4 video with missing "moov atom"   is there way  ffmpeg can do the same thing?
[04:02:00 CEST] Last message repeated 1 time(s).
[04:17:00 CEST] Last message repeated 1 time(s).
[05:26:32 CEST] <TikityTik> How inaccurate is input seeking vs output seeking? Is there a way to really speed up output seeking?
[07:12:50 CEST] <Prest0o> hi, anyone know why ffmpeg doesnt read the file when i use the * ?
[07:13:28 CEST] <Prest0o> since sometimes the jpg file might have a diferent name, i used the command *.jpg so it take the jpg whatever name it has, but doesn't read it, it want a static path
[07:18:31 CEST] <Prest0o> [image2 @ 000000324fc9a440] Could find no file with path '*.jpg' and index in the range 0-4
[07:18:31 CEST] <Prest0o> *.jpg: No such file or directory
[07:34:43 CEST] <furq> Prest0o: add -pattern_type glob before -i
[07:35:43 CEST] <Prest0o> doenst work
[07:36:02 CEST] <Prest0o> aparently this just works on linux
[08:07:29 CEST] <irwiss> I'm playing a stream from a network cam, using the computer clock and frame pts converted to timestamp (shifted by snapshot of computer clock i took when i started playing) i'm deciding when to display the frame and pull next one. After a while i'm drifting pretty far off the mark, up to 30 seconds so i'm guessing the 2 clocks drift apart, as i und
[08:07:29 CEST] <irwiss> erstand it the camera clock is likely just a cheap oscillator alone, so it's probably drifting about. How can i deal with this? Do i adjust my timestamp snapshot once i see it drifts too far? Can i rely on packet PTS value to be somewhat related to what i'll get on frame pts?
[09:22:05 CEST] <__raven__> how to slow down a video by ~speed/4 with interpolation but avoiding morphing of cuts?
[09:23:40 CEST] <Prest0o> noo idea
[09:23:41 CEST] <furq> if you mean with minterpolate then all you can really do is adjust scd_threshold
[10:03:08 CEST] <hannes_> Prest0o: what is your pattern and what is an example filename?
[10:03:14 CEST] <hannes_> " and index in the range 0-4"!
[10:04:32 CEST] <Prest0o> hannes_
[10:04:34 CEST] <Prest0o> ffmpeg -loop 1 -y -i image.jpg -f concat -i listamp3.txt -shortest -acodec copy -vcodec h264 -vf scale=320:240 result.avi
[10:05:07 CEST] <Prest0o> if somethinge the image i will use has another name, i want it takes the image without me losing time renaming it to image.jpg
[10:05:16 CEST] <Prest0o> but if i use -i *.jpg
[10:05:21 CEST] <Prest0o> it doesnt read it
[10:07:45 CEST] <hannes_> so there is just one image in that directory?
[10:08:23 CEST] <Prest0o> yes
[10:08:38 CEST] <Prest0o> i just concatenate mp3 files to 1 image and create a video
[10:09:59 CEST] <hannes_> hm, no idea about windows...
[10:10:09 CEST] <Prest0o> okey no problem, thanks anyway
[10:10:38 CEST] <hannes_> if you can, maybe write a loop (that will just run once) like this (linux shell): for image in *.jpg; do ffmpeg ... -i "${image}" ... ; done
[10:11:00 CEST] <hannes_> that would make sure to present ffmpeg with the actual filename instead of the verbatim *
[10:11:12 CEST] <hannes_> maybe try quoting the * in "" or ''
[10:11:38 CEST] <hannes_> https://superuser.com/questions/460598/is-there-any-way-to-get-the-windows-cmd-shell-to-expand-wildcard-paths
[10:28:29 CEST] <Prest0o> hannes_ maybe you have an idea about another problem i have, when i have 14 files, named from 1.mp3 - 2.mp3 .... to 14.mp3
[10:28:32 CEST] <Prest0o> and i list them
[10:28:50 CEST] <Prest0o> with cmd: dir *.mp3
[10:29:10 CEST] <Prest0o> it always shows 1.mp3 then 11.mp3 then 12.mp3 then 14.mp3 then 2.mp3...
[10:29:50 CEST] <Prest0o> why so stupid to think 11 comes after 1 when there is a 2 ?
[10:38:18 CEST] <TheAMM> '1' < '.'
[10:38:35 CEST] <TheAMM> zero-pad your numbers
[10:40:26 CEST] <Prest0o> yes TheAMM
[10:40:34 CEST] <Prest0o> https://pastebin.com/UMjjU9G0 but how i add the 0 before the 1 ?
[10:40:42 CEST] <Prest0o> i tried to put "00" but doesnt work
[10:44:25 CEST] <TheAMM> dunno about windows
[11:49:28 CEST] <__raven__> furq: ok tnx i will try
[16:30:19 CEST] <superware> is it possible sws_scale randomly takes between 1000 and 4000ms? all frames has the same size...
[16:54:16 CEST] <BtbN> for a single frame?
[16:55:48 CEST] <superware> yep
[16:56:06 CEST] <BtbN> Must be very large res or super slow CPU
[16:56:28 CEST] <superware> 1280x720 AV_PIX_FMT_YUVJ420P to AV_PIX_FMT_BGR24 (no scaling)
[16:56:45 CEST] <superware> i5, not slow
[16:58:00 CEST] <kepstin> yuv to rgb conversion on that cpu for a 720p video would be expected to be only a few ms at worst
[16:58:04 CEST] <kepstin> not 1-4 seconds
[16:58:34 CEST] <kepstin> are you sure you don't have milliseconds and microseconds mixed up?
[17:13:37 CEST] <superware> kepstin: my bad! I meant 1-4ms
[17:13:51 CEST] <superware> :)
[17:14:31 CEST] <kepstin> right, variations in the small number of ms might be down to OS scheduling or cache behaviour :/
[17:14:52 CEST] <kepstin> hard to say what's up there
[17:27:15 CEST] <superware> 4ms sounds like a lot
[17:27:48 CEST] <kepstin> yeah
[17:28:53 CEST] <kepstin> are you re-using a swscale context for multiple frames or setting it up per frame?
[17:29:27 CEST] <superware> yep
[17:29:47 CEST] <kepstin> some conversions require runtime setup for the scaler that only needs to be done once if you can re-use the context. dunno if this is one of them.
[17:29:59 CEST] <superware> reusing
[17:30:38 CEST] <superware> thanks
[17:31:33 CEST] <kepstin> hmm. on an i5, one thing to consider is that if the cpu isn't fully loaded, the cpu will dynamically reduce clock speed to the minimum necessary for the running apps.
[17:32:03 CEST] <kepstin> (or, in some cases, an overloaded cpu can thermal throttle)
[17:32:24 CEST] <DHE> I think that's more a linux (?) CPU scaling governor thing
[17:37:52 CEST] <kepstin> hmm, newer intel cpus the default governor tells the cpu's power controller to adjust speed automatically. you can adjust the policy a bit via the linux driver, or load a different governor to take software control.
[17:41:41 CEST] <kepstin> ryzen cpus (with default power management when not overclocked) are even weirder, because they run a software governor and hardware governor simultaneously - software requests one of three speed tiers, then the hardware picks a speed within a fairly large range within that tier.
[17:42:10 CEST] <DHE> yay turbo boost!
[17:44:06 CEST] <kepstin> on my r7 1700, setting the speed to 3000Mhz (the highest tier) via the cpufreq interface lets the cpu pick any speed between about 2.8 and 3.75ghz depending on load and cooling.
[21:22:52 CEST] <TikityTik> How inaccurate is input seeking vs output seeking? Is there a way to really speed up output seeking?
[21:23:16 CEST] <TheAMM> Both should be as accurate if the file is proper
[21:23:25 CEST] <TikityTik> Also, why is it that when ffmpeg prints its encoding progress that it keeps doing new lines when my terminal windows isn't wide enough?
[21:23:40 CEST] <furq> because your terminal isn't wide enough
[21:23:54 CEST] <TheAMM> Output seeking decodes everything and discards it until it reaches the timestamp - there's no way to speed it up, because you have to decode to decode
[21:23:55 CEST] <TikityTik> is there a way to stop the newlines from happening when the terminal is narrow?
[21:23:59 CEST] <furq> no
[21:24:28 CEST] <TikityTik> can you make the progress show less info then so it doesn't happen? like just showing filesize, frames, and length/duration?
[21:25:11 CEST] <furq> if you want the technical reason it's because the progressbar works by sending ansi escape codes
[21:25:18 CEST] <furq> presumably clear to start of line and carriage return
[21:25:32 CEST] <furq> so if your terminal needs a second line then those will only clear the second line
[21:26:01 CEST] <TikityTik> i guess i could parse the output through bash somehow
[21:26:15 CEST] <TikityTik> thanks for the info guys
[21:40:58 CEST] <wallace_mu> hi all, any idea why this command failed on `Option 'force_style' not found`
[21:40:59 CEST] <wallace_mu> ffmpeg -loop 1 -y -i 1.png -i 1.mp3 -vf  "ass=my_subtitles_edited.ass:force_style='Fontsize=24,PrimaryColour=&H0000ff&'" -shortest  -preset veryfast test2.mp4
[21:42:26 CEST] <wallace_mu> seems ass filter doesn't support force_style ?
[21:45:28 CEST] <TikityTik> I can't believe this bug is still open. https://trac.ffmpeg.org/ticket/2067
[21:45:45 CEST] <TikityTik> i might even end up fixing this myself
[21:45:58 CEST] <wallace_mu> a different problem is no matter what fontname i specify in the ass file, when i run the command, ffmpeg just choose Arial
[21:46:17 CEST] <JEEB> wallace_mu: sounds like you don't have a fontconfig config or something?
[21:46:31 CEST] <JEEB> since I think the subtitles/ass filter utilizes fontconfig to figure out fonts (and I think libass as well)
[21:46:44 CEST] <JEEB> actually, libass itself should be possible to build with various back-ends
[21:46:50 CEST] <JEEB> fontconfig being one of htem
[21:46:56 CEST] <JEEB> for windows it can use the system font cache
[21:47:15 CEST] <JEEB> TikityTik: yea, subtitles are still the only type of stream which doesn't utilize AVFrames :)
[21:47:31 CEST] <JEEB> one of the devs has been developing a switch to AVFrames for subtitles
[21:47:39 CEST] <JEEB> but it's a big n' painful road
[21:47:50 CEST] <TikityTik> lol
[21:47:57 CEST] <wallace_mu> JEEB: thanks for replying, just know `fontconfig ` this term you mentioned, i'll search how to use it
[23:01:47 CEST] <wallace_mu> JEEB: one more question, do i have to specify fontconfig, i am curious why the font i set in ass file not get picked up by ffmpeg
[23:06:41 CEST] <wallace_mu> looks like i can speficy \fn before each text, it would apply, thx anyway
[00:00:00 CEST] --- Thu May  2 2019


More information about the Ffmpeg-devel-irc mailing list