[Ffmpeg-devel-irc] ffmpeg.log.20151209

[burek]

[00:20:45 CET] <debdog> I have a video which is a tad overexposed, see: http://beggabaur.rocks/daten/shot0003.png
[00:20:45 CET] <debdog> now I wonder what would be the best way to get the grainy/noisy stuff away. by making it darker or using some grain/noise filter (and if so, which one?)?
[00:21:05 CET] <debdog> just to say, I am a total noob when it comes to that video stuff
[00:21:48 CET] <debdog> reading http://ffmpeg.org/ffmpeg-filters.html and found one grain and some noise related filters
[00:32:41 CET] <durandal_1707> depend what kind of noise..
[00:34:03 CET] <debdog> hum, dunno how to explain. noise from overexposed filming. like seen on that shot
[00:34:16 CET] <debdog> which actually is from that said clip
[00:38:04 CET] <debdog> hmm, I think I just try http://ffmpeg.org/ffmpeg-filters.html#toc-colorlevels first
[01:24:51 CET] <nullp0inter> i am trying to resize my videos to be mobile friendly however my ffmpeg command does not seem to be doing what i need it to do: http://pastebin.com/FK9nZfw3 anyone have any ideas?
[01:25:18 CET] <DHE> output file must come last
[01:25:21 CET] <nullp0inter> also, it seems that the audio is choppy and malformed when played on an iphone, no idea why
[01:26:14 CET] <nullp0inter> DHE, ty i will try that now
[01:53:00 CET] <nullp0inter> DHE i did that, but now im getting this error: http://pastebin.com/Vq6S7Raw
[01:53:53 CET] <nullp0inter> the moov part is a second pass
[01:56:55 CET] <jasom> debdog: the 3d denoise is a good place to start with denoising
[01:57:36 CET] <jasom> debdog: and that looks properly exposed; the only clipping is on the spots, which probably should clip
[01:57:54 CET] <nullp0inter> DHE ignore that i had the wrong lines
[01:58:53 CET] <jasom> debdog: https://ffmpeg.org/ffmpeg-filters.html#hqdn3d-1 <-- documentation there
[01:59:47 CET] <nullp0inter> can anyone tell me why i would get this error: Error while opening encoder for output stream #0:0 - maybe incorrect parameters such as bit_rate, rate, width or height
[02:00:04 CET] <jasom> debdog: you may want to tune the chroma parameters a bit as there appears to be a lot of chroma noise in the shot you posted
[02:02:10 CET] <nullp0inter> this is my command producing that error: http://pastebin.com/fLnTMemT
[02:02:36 CET] <jasom> debdog: maybe 3:4:4:6 as a good starting point for the options
[02:31:08 CET] <cluelessperson> with ffserver, can you just disable the audio stream?
[03:00:24 CET] <relaxed> nullp0inter: pastebin.com your ffmpeg_vid.log
[03:01:38 CET] <relaxed> nullp0inter: you have -c:a copy and -c:a libfdk_aac
[03:02:04 CET] <pzich> there's also a good chance your scale is causing the width to be an odd number, and you need it to be a multiple of of two for 264
[03:04:59 CET] <relaxed> true, use scale=-2:480
[03:49:32 CET] <nedrocks> In order to fix an issue I have with audio streams not starting at the proper times (causes audio sync issues) I need to fill the first block of silence detected with silencedetect filter with.. something. Unfortunately Im having issues grabbing the first silent timestamp from the output of the silencedetect filter (see the marked line here: http://pastebin.com/EingS6EK). Can anyone give me some advice as to how to get this range more
[03:49:33 CET] <nedrocks> easily than parsing output from stderr?
[04:07:49 CET] <waressearcher2> nedrocks: hallo
[05:18:14 CET] <cluelessperson> Hi all.  I'm trying to use ffserver, but the problem is I get this
[05:18:14 CET] <cluelessperson> 1.ffm: Pid 8954 exited with status 256 after 0 seconds
[05:18:32 CET] <cluelessperson> pretty much immediately, I think it may be a problem with the ffmpeg line
[05:18:51 CET] <cluelessperson> Launch ffmpeg -loop 1 -i test.mpg
[05:19:04 CET] <cluelessperson> does that look right?
[05:27:39 CET] <Prelude_Zzzzz> hey.. having trouble with vdpau and startx can anyone help?
[05:28:47 CET] <cluelessperson> Here's the full command and output.   http://dpaste.com/0EVHGSQ
[05:28:51 CET] <cluelessperson> not sure what it's asking for
[05:29:11 CET] <cluelessperson> [libvpx @ 0xecd900] CQ level 0 must be between minimum and maximum quantizer value (2-31)
[05:29:12 CET] <cluelessperson>  ?
[05:54:36 CET] <cluelessperson> at http://cluelessperson.com/stream/   You can see I have ffserver running, listening for 1.ffm     at http://cluelesspreson.com/stream/1.webm    You can see there's no output.
[05:59:53 CET] <waressearcher2> cluelessperson: hallo
[06:00:03 CET] <cluelessperson> waressearcher2, hi there
[06:00:10 CET] <waressearcher2> cluelessperson: wie geht's ?
[06:00:24 CET] <cluelessperson> waressearcher2, sorry?
[06:00:37 CET] <cluelessperson> german?
[06:01:00 CET] <waressearcher2> deutsch
[06:02:26 CET] <cluelessperson> waressearcher2, How are you?
[06:11:57 CET] <Admin_234234> hey question.. anyone know if vdpau can run without actually startX running ?
[06:12:09 CET] <Admin_234234> like using just the server CLI or whatever
[06:12:12 CET] <Admin_234234> need it for decoding
[07:55:19 CET] <cluelessperson> okay, so I've managed to get somethig streaming, but it only works if the webm/browser is already open, and THEN you start the stream
[08:06:38 CET] <waressearcher2> cluelessperson: es geht's
[08:07:02 CET] <cluelessperson> waressearcher2, :D   https://cluelessperson.com/stream/1.webm
[08:52:53 CET] <waressearcher2> cluelessperson: http://www.youtube.com/watch?v=dQw4w9WgXcQ
[11:41:05 CET] <termos> I'm trying to create a 30 minutes delayed stream with FFmpeg and would like the possibility to resume if something happened without having to buffer up the 30 minutes delay on restart. Is there such a feature in FFmpeg?
[13:52:01 CET] <dantti> does somenone knows how a m3u8 playlist should look like if I have an infinite .ts file?
[14:05:14 CET] <DHE> I don't think that's supported by HLS
[14:11:05 CET] <dantti> DHE: doesn't HLS stands for Live stream? how can it be live otherwise?
[14:11:37 CET] <flux> by using multiple files?
[14:12:56 CET] <flux> I don't know however if ffmpeg is suitable for creating a live HLS..
[14:13:51 CET] <dantti> flux: but if you have multiple files your playlist has an end no?
[14:14:23 CET] <DHE> typically you delete the older videos after some period of time. there's an option for that under -hls_flags
[14:14:24 CET] <flux> I think the purpose is to update the playlist
[14:14:27 CET] <DHE> I'm doing live HLS. right now
[14:16:57 CET] <dantti> right, atm I have my own http stuff that writes the mpeg-ts directly to the socket not to write files
[14:17:10 CET] <dantti> so I guess I'd need to have a playlist with fake .ts files
[14:19:09 CET] <flux> I understand one idea behind these chunked representations is that they are easier to cache (ie. in a content delivery network) than big streaming files
[14:20:57 CET] <flux> does ffmpeg have live DASH streaming support? seems like it has some required parts but not a 'complete' solution
[14:21:21 CET] <DHE> having used HLS quite a bit there's more to it than that. clients seem to always download a whole chunk before they start playing - they don't do live playing from an HTTP transfer. so smaller chunks (to a point) are encouraged for faster playback initialization
[14:21:51 CET] <flux> I guess that's something a player can choose to do otherwise
[14:22:11 CET] <flux> but the point in downloading a full chunk before playing might be that the user is able to determine the connection bitrate, so they know which fragment to download next
[14:22:44 CET] <DHE> that seems to be the trend I've seen. so I chose 6 seconds as my starting point for chunking size
[14:23:01 CET] <DHE> dantti: ffmpeg has `-f hls` where you specify the .m3u8 file as the write target
[14:23:26 CET] <flux> https://github.com/sannies/rtp2dash with ffmpeg seems to be pretty close to a complete solution. but.. java :)
[14:34:50 CET] <dantti> DHE: how do you mean m3u8 as a write target? I want to avoid writing to files as I think it might damage internal storage
[14:34:59 CET] <DHE> wat?
[14:35:14 CET] <dantti> also I will only support 1client at time for vieweing the tablet's camera
[14:35:21 CET] <DHE> sounds more like you want live RTSP streaming
[14:35:43 CET] <dantti> well hls is more easily supported in ios and android as player
[14:37:19 CET] <dantti> at least afaik
[14:37:53 CET] <flux> dantti, you have flash internal storage?
[14:38:06 CET] <flux> dantti, maybe you can use ramdisk for storing the file
[14:39:27 CET] <dantti> flux: last time I checked a ramdisk requires root, also I could also store the files in a buffer on the app
[14:40:04 CET] <dantti> but writting to a socket doesn't seems like a bad decision after all it's live  :P
[14:40:19 CET] <flux> ramdisk on what?
[14:40:36 CET] <flux> what kind of device do you have that uses flash but you don't have access to root?
[14:40:39 CET] <dantti> android
[14:40:50 CET] <dantti> those cheap android tablets
[14:40:55 CET] <flux> how about /tmp of /var/run/tmp? I don't know if they are limited.
[14:41:08 CET] <flux> (actually just /var/run)
[14:41:17 CET] <flux> they might be tmpfs
[14:41:48 CET] <flux> alternatively have your own program that keeps the data in the memory and have the web server query the data from it..
[14:41:49 CET] <dantti> android is such a jungle that I think the vendor can also change that..
[14:42:03 CET] <dantti> well that's the second idea
[14:43:17 CET] <dantti> but then the m3u8 should have say 5 segments, and does the player would loo it?
[14:45:20 CET] <flux> loop it? no, I don't expect it to loop it.
[14:45:37 CET] <flux> https://developer.apple.com/library/ios/technotes/tn2288/_index.html has examples of what the .m3u8 looks like while live streaming
[14:47:05 CET] <dantti> flux: thanks I'll have a look
[14:48:52 CET] <dantti> flux: so, after the end of the playlist the player fetches the m3u8 agains and it has different sequenceX.ts  files right?
[14:48:59 CET] <flux> yes
[14:49:17 CET] <dantti> ok, now it makes a bit more sense :)
[16:23:23 CET] <debdog> jasom: thanks a lot! it wasn't easy and I needed quite some approaches. the original shot again: http://beggabaur.rocks/daten/shot0003.png and the result http://beggabaur.rocks/daten/shot0001.png
[16:23:23 CET] <debdog> I wasn't able to get rid of the noise entirely, then the other parts of the vid would have been bad. but it looks very good now!
[16:23:23 CET] <debdog> I could PM the URL to the entire video later, if you're interested. it's a laser show at Europa Park, Germany. we intend to upload it to vimeo or such, it'll be public then.
[16:35:34 CET] <eightfold> im trying to create a video encoded with h.264 that will be playable in safari
[16:35:42 CET] <eightfold> im not succeeding
[16:36:28 CET] <eightfold> via the html5 video element...
[16:36:41 CET] <eightfold> what ive done: ffmpeg -f gif -i input.gif -vcodec libx264 -level 21 -refs 2 -b 1100K -bt 1100K -threads 0 output.mp4
[16:36:49 CET] <eightfold> does safari need a mov container?
[16:40:59 CET] <kepstin> in safari on what os? this differs between desktop and mobile, I think
[16:41:55 CET] <kepstin> I think safari on mobile requires hls for video streaming
[16:42:08 CET] <eightfold> ive just tried on os x
[16:42:19 CET] <eightfold> could it be a mime type problem perhaps?
[16:47:10 CET] <TD-Linux> eightfold, you can't live stream a mp4 container
[16:52:27 CET] <kepstin> eightfold: assuming you're not live streaming (which it doesn't look like you are), you could try adding "-movflags faststart", which might help the browser load the video over http...
[16:52:51 CET] <kepstin> other than that, start looking in the browser web inspector console and network tab for issues...
[16:53:49 CET] <eightfold> no live streaming
[16:57:04 CET] <durandal_1707> debdog: what you used?
[16:58:24 CET] <cluelessperson> http://cluelessperson.com/stream/
[16:58:35 CET] <cluelessperson> I got it working. :)
[17:00:07 CET] <eightfold> hmm, did not help
[17:00:14 CET] <eightfold> the -movflags faststart
[17:00:22 CET] <eightfold> anything else in my line looking unnecessary
[17:00:29 CET] <eightfold> cut n paste job, of course
[17:17:31 CET] <eightfold> this is a mystery, nothing can get a h.264 encoded video to play in safari
[17:18:12 CET] <eightfold> using <video autoplay loop muted><source type="video/mp4" src="test.mp4"><source type="video/webm" src="test.webm"></video>
[17:23:13 CET] <debdog> durandal_1707: I've tried a lot of stuff, cannot find the command line atm. but the fines clip only was compressed, which made it a tad darker, and reduced gamma for about 10 %
[17:24:14 CET] <debdog> s/fines/final/
[17:50:23 CET] <qubodup> hello, can I record a screen capture at 1 frame per second but have the output file play it at 30fps without additional steps? http://superuser.com/questions/1010940/
[17:59:44 CET] <c_14> qubodup: use -framerate 1 as an input option and -r 30 as an output option
[18:01:07 CET] <dantti> on a new hls segment should I mark AVFrame->key_frame = 1?
[18:06:06 CET] <qubodup> c_14: unfortunately this doesn't work. the playback speed of the resulting video is 1fps
[18:07:31 CET] <qubodup> c_14: ffmpeg -framerate 1 -f x11grab -s 1280,720 -i :0.0+0,0 -vcodec libx264 -crf 0 -preset ultrafast -r 30 out.mkv
[18:07:39 CET] <popara> Hello Guys. One question
[18:07:55 CET] <popara> How i can make FFmpeg NOT to edit any timestamp from the input video?
[18:08:27 CET] <popara> The problem is that if an INPUT loose sound, but later come again, the FFmpeg is keep restreaming it without sound because it has mess up with the timestamps (i think)
[18:33:57 CET] <c_14> you can try copyts and/or vsync 0
[18:35:18 CET] <c_14> qubodup: use -vf setpts=N/(30*TB)
[18:45:07 CET] <DHE> <dantti> on a new hls segment should I mark AVFrame->key_frame = 1?  # yes. the hls muxer will split the video on keyframes normally if you use that.
[18:51:03 CET] <excalibr> Not sure if this is a proper place to ask but anyone has any clue how gfy animated images are created? Can use you use ffmpeg to create gfy?
[18:51:56 CET] <furq> excalibr: ffmpeg -i source.gif dest.mp4
[18:52:05 CET] <furq> or dest.webm
[18:53:47 CET] <excalibr> furq, gfy format is essentially that? But how optimization on non changing region of the images is done?
[18:54:37 CET] <excalibr> Cause I heard or read somewhere that optimization is what made gfy much smaller and better than regular gif
[18:54:59 CET] <dantti> DHE: is the hls server api available for apps to use? tho I think being able to reconfigure the camera resolution depending on the connection might be easier if I implement that myself I guess
[18:58:09 CET] <prelude2004c> crazy disconnects :(
[18:58:19 CET] <prelude2004c> did anyone reply to something i did not see ?
[18:58:30 CET] <prelude2004c> icechat crashed :(
[19:00:12 CET] <[SLiM]D12> hey i have a dumb question
[19:00:33 CET] <[SLiM]D12> i know to download a video from m3u8 it's generally "ffmpeg -i "master.m3u8" -c copy output.ts
[19:00:33 CET] <prelude2004c> slim , no such thing... you haven't read mine :P haha
[19:00:44 CET] <[SLiM]D12> is there a way to define a specific bitrate of the file i want from the m3u8?
[19:00:51 CET] <[SLiM]D12> like not the best quality, but 2nd best for example?
[19:01:20 CET] <prelude2004c> slim, look inside that m3u8 .. if it has multi bit rates generally you can just go to that m3u8 directly for the desired bit rate
[19:01:33 CET] <[SLiM]D12> yeah i tried that, doesn't work
[19:01:35 CET] <prelude2004c> there is a master M3u8 , then sub m3u8s ( one for each bit rate ) and then the .ts files listed
[19:01:42 CET] <[SLiM]D12> gives me a 403 forbidden
[19:01:48 CET] <[SLiM]D12> even though i can download the single ts files within the m3u8
[19:01:59 CET] <[SLiM]D12> i get a forbidden when pointing to any m3u8 other than the master
[19:02:22 CET] <prelude2004c> sounds like a permissions issue ... odd really
[19:02:29 CET] <furq> excalibr: the optimisation is that it's a video
[19:02:31 CET] <[SLiM]D12> i know you can do it in livestreamer, and that works, but there are 2 720p streams, and livestreamer's picking the lower quality one
[19:02:38 CET] <[SLiM]D12> i want to pick the higher quality 720p stream
[19:02:46 CET] <furq> gif doesn't do inter-frame prediction
[19:17:02 CET] <prelude2004c> so any suggestions if not vdpau .. basically i want to lower cpu decoding and push it to GPU in order to do the transcosing
[19:17:24 CET] <prelude2004c> encoding is already working with nvenc so thats good
[19:18:31 CET] <prelude2004c> the probelm now is the decoding part that is still using cpu's to decode and then pass to GPU for encode. i want o use up the CPU decoding capabilies.. any suggestions or pointing in the right direction would be helpful
[19:25:17 CET] <lindenk> Hey, so I have a video that was recorded on a phone and, when played in mpv and vlc, it works fine, however, if I attempt to play it in windows media player, it's rotated 90 degrees counterclockwise. I attempted to rotate it with ffmpeg, however that just seemed to turn it 90 clockwise regardless of what I played it with. I'm guessing this is a problem with the metadata associated with the video, and I've
[19:25:19 CET] <lindenk> tried -metadata:s:v, however this made the video unplayable in wmp, but still right side up with mpv and vlc. Any ideas?
[19:25:48 CET] <lindenk> the video is encoded as an mp4
[21:05:33 CET] <floyddotnet> hello, i try to find out the right settings to convert a 720p h264 source to h265 720p with less quality losses. so i try 2-pass vbr 4 and also 1-pass crf 21 but both are very washy. take a look at http://screenshotcomparison.com/comparison/153693 and http://screenshotcomparison.com/comparison/153694 .. can you give me some hints what i'm doing wrong?
[21:24:33 CET] <BtbN> floyddotnet, you won't magicaly make it look better by re-encoding. Only worse.
[21:25:21 CET] <floyddotnet> not better, but compareable?
[21:25:35 CET] <floyddotnet> same quality less file size
[21:26:00 CET] <fritsch> yeah and 4 times the power consumption while decoding ...
[21:26:04 CET] <fritsch> hard disk space is cheap
[21:26:14 CET] <fritsch> (not mentioning the insane slow hevc encoding)
[21:28:04 CET] <floyddotnet> this is not the question ;) btw the target hardware has h265 4k at 60fps hw-decoder
[21:28:33 CET] <fritsch> whatever you wast energy while transcoding
[21:28:36 CET] <fritsch> and reduce quality
[21:29:00 CET] <fritsch> if you are at it encode it 10 bits ... as it "compresses better" (TM) (R)
[21:31:00 CET] <floyddotnet> 10 bits, yes
[21:32:48 CET] <fritsch> yeah, no comment
[21:33:00 CET] <fritsch> you were a h264 hi10p user, too right?
[21:33:24 CET] <floyddotnet> oh .. you mean the source
[21:33:46 CET] <fritsch> no
[21:33:48 CET] <floyddotnet> i'm not sure .. Decodiertes Format: Planar 4:2:0 YUV
[21:34:01 CET] <fritsch> there is no source with 10 bit for bluray
[21:34:32 CET] <floyddotnet> no bluray
[21:34:49 CET] <fritsch> there is no mainline native content in 10 bit as of now
[21:35:06 CET] <fritsch> but coming soon
[21:36:07 CET] <floyddotnet> ffprobe-output for the source: Stream #0:0(eng): Video: h264 (High), yuv420p, 1280x536 [SAR 1:1 DAR 160:67], 23.98 fps, 23.98 tbr, 1k tbn, 47.95 tbc (default)
[21:36:28 CET] <fritsch> keep it as it
[21:36:31 CET] <fritsch> it won't get better
[21:37:18 CET] <floyddotnet> is there no way to convert it to h265 with near equal quality?
[21:37:41 CET] <fritsch> it makes no sense ...
[21:38:01 CET] <floyddotnet> this is not the question ;)
[21:38:07 CET] <fritsch> and at this resolution it also makes no sense
[21:38:45 CET] <floyddotnet> why .. h265 should resulting in a smaller file
[21:39:11 CET] <floyddotnet> estimated half file size at same qualitiy if i'm understand it right
[21:39:44 CET] <fritsch> :-(
[21:39:47 CET] <fritsch> yeah - read up again
[21:39:55 CET] <fritsch> or don't watch too much commercials
[21:40:17 CET] <floyddotnet> nobody say's something about commercials
[21:40:29 CET] <fritsch> argh - your information is wrong
[21:40:38 CET] <fritsch> you did watch too much commercials
[21:40:46 CET] <fritsch> about hevc
[21:41:00 CET] <floyddotnet> oh i understand .. sorry :)
[23:14:40 CET] <qubodup> c_14:  -vf setpts=N/(30*TB) while recording doesn't seem to work
[23:41:04 CET] <prelude2004c> hey guys.. stumbled upon " NVCUVID " ... anyone know i could use that with ffmpeg ? i can't seem to find any good documentation online
[23:41:30 CET] <tp_> its like 30% savings for h265, but reencoding a h264 video for that reason is just wrong
[23:44:55 CET] <prelude2004c> tp that for me ?
[23:51:23 CET] <tp_> nope :P
[00:00:00 CET] --- Thu Dec 10 2015