[Ffmpeg-devel-irc] ffmpeg.log.20180923

[burek]

[00:04:13 CEST] <WithinRafael> poutine: I'm streaming radio audio onto my local network via udp/rtp and would like to supplement that with some time-aware metadata, am flexible with container formats, metadata format, etc.
[00:05:26 CEST] <WithinRafael> poutine: (metadata, for example, like the transmitter source, signal strength, etc.)
[00:20:57 CEST] <kiroma> Hey, is there a filter that would let me normalize saturation of a video?
[00:27:10 CEST] <kiroma> Or optionally a filter that will show me by how much will I have to boost the saturation?
[00:28:37 CEST] <Harzilein> hi
[00:33:29 CEST] <matteo> hi all people
[00:34:04 CEST] <matteo> quick question, I have an integrated Intel GPU, can I use it to encode h.264 video?
[00:36:03 CEST] <kiroma> Yes, you can use vaapi
[00:37:06 CEST] <kiroma> https://trac.ffmpeg.org/wiki/Hardware/VAAPI
[00:53:01 CEST] <kiroma> Is it possible that an nvidia GPU would be able to encode a H264 vid but lack the capabilities to decode it?
[01:24:19 CEST] <matteo> kiroma: thanks
[01:24:26 CEST] <matteo> is it much faster than a CPU?
[01:24:43 CEST] <matteo> I guess that intel GPU aren't much powerful
[01:30:03 CEST] <kiroma> matteo: It could be, Intel's iGPUs are suprisingly good with stuff like compute/geometry shaders so you may notice a significant speed increase.
[01:30:26 CEST] <matteo> really? good to know then!
[01:30:37 CEST] <matteo> do I need intel media sdk or wathever?
[01:30:39 CEST] <tdr> yeah depending on what youre asking them to do, theyre alright
[01:30:52 CEST] <matteo> I've found this http://mediasdk.intel.com/
[01:31:01 CEST] <kiroma> NO DO NOT INSTALL THE SDK
[01:31:08 CEST] <matteo> ok
[01:31:16 CEST] <kiroma> Under no circumstances should you do that, it breaks the entire system
[01:31:21 CEST] <kiroma> I made that mistake once
[01:31:28 CEST] <matteo> lol, ok. thanks
[01:32:29 CEST] <kiroma> Afaik there are no special requirements for building ffmpeg with vaapi so it should work out of the box by default.
[01:33:43 CEST] <kiroma> Follow the guide in the link I gave you and it should work just fine.
[01:36:18 CEST] <matteo> great. thanks a lot, will do
[01:40:27 CEST] <matteo> kiroma: what is the preferred between libmfx, libva with i965 and libva with iHD?
[01:41:01 CEST] <matteo> I have an Intel HD 530
[01:41:19 CEST] <matteo> the one embedded in the 6700K
[01:42:34 CEST] <kiroma> I don't know, someone else will have to answer that.
[01:49:35 CEST] <matteo> thanks anyway, I'll try i965 first
[01:49:37 CEST] <matteo> gnite
[03:06:39 CEST] <boblamont> When I use ffmpeg -f concat -safe 0 -i does it reencode the source files?
[03:08:11 CEST] <BtbN> If you tell it to
[03:08:52 CEST] <boblamont> the rest of what I use is  $DIR/inputs.txt -c copy $DIR/output.mp3
[03:08:57 CEST] <BtbN> Or rather, don't tell it not to.
[03:09:05 CEST] <BtbN> That's just remuxing then
[03:11:08 CEST] <boblamont> I run into problems if the sample rate is different (i.e. 48 instead of 41) or if a file is mono and the rest are stereo, but it seems fine with different bitrates... why is it fine with different bitrates?
[03:12:07 CEST] <BtbN> Because the bitrate doesn't matter for the decoder
[03:12:13 CEST] <BtbN> randomly changing output parameters do
[03:12:32 CEST] <BtbN> you'll have to transcode at least some parts of it to concat it
[03:13:39 CEST] <boblamont> yeah, I fix any sample rate or channel number problems (I don't always catch the mono files, but it's pretty obvious when I get to part that sounds like the chipmunks
[03:16:36 CEST] <boblamont> is there any harm in using the same command to combine files I made using the command before (combining the combined files)?
[03:16:50 CEST] <BtbN> if they're compatible, no
[03:18:47 CEST] <boblamont> right, I think there's generally a baseline of stereo 41.1khz, though with different bitrates, so most of the first batch of combined files should be consistent
[03:19:26 CEST] <BtbN> you can just transcode the files that aren't to that format, and then freely concat
[03:20:06 CEST] <boblamont> is there quality loss anywhere?
[03:20:14 CEST] <BtbN> everytime you transcode
[03:20:26 CEST] <boblamont> but not when I concat?
[03:20:28 CEST] <BtbN> no
[03:20:53 CEST] <BtbN> unless you transcode at the same time
[03:21:25 CEST] <boblamont> I'm not in the command I pasted, right?
[03:22:29 CEST] <boblamont> ffmpeg -f concat -safe 0 -i $DIR/inputs.txt -c copy $DIR/output.mp3
[03:24:06 CEST] <boblamont> you mentioned "not telling it not to"
[04:56:04 CEST] <Matador> weee
[08:39:56 CEST] <infantilo> hi
[08:40:46 CEST] <infantilo> anyone here that could give me advice with using ANULLSINK causing conversion to run vorever, even when seeting time? THX for any help.
[11:31:42 CEST] <linuxmaster> did anyone know something about pitime?
[11:31:55 CEST] <linuxmaster> ptime i mean
[11:55:06 CEST] <poutine> the answer to my caption question earlier was to just use "X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:0" in the WebVTT files, and give timestamps of start PTS / time base for the WebVTT timestamps
[15:37:27 CEST] <adrian_1908> Hello. I'm trying to convert an m4b file into re-encoded mp3 files, split by chapter. I don't think ffmpeg alone can do this, but it certainly can see (print) the metadata of the m4b file. Could anyone show me how to export that metadata to a file?
[15:39:33 CEST] <adrian_1908> Ah, now that I make the effort ask i actually stumble upon a solution: https://stackoverflow.com/questions/9464617/retrieving-and-saving-media-metadata-using-ffmpeg
[15:39:35 CEST] <adrian_1908> Sorry!
[17:37:14 CEST] <Bombo> is there a command to get the encoding options of a x264 stream? i see x264 version and some options in hexeditor ;) it doesn't seem to be a normal metadata/tag
[18:05:22 CEST] <cslcm> Hi all - I was wondering if anyone has played with implementing a RAISR upsampling filter for ffmpeg? Seems that there's an opportunity to accelerate it with tensor cores in nvidia's new gpus
[18:05:32 CEST] <cslcm> or some other machine learning based upsampler
[18:07:49 CEST] <durandal_1707> is it useful?
[18:08:44 CEST] <cslcm> well, i've used the python implementation to upsample DVD-resolution video. Upsampling to 1080p works extremely well, and results in a pretty stunning image. Upsampling to 4k introduces a lot of artifacts.
[18:08:49 CEST] <cslcm> so it's useful to a point
[18:09:49 CEST] <cslcm> https://ai.googleblog.com/2016/11/enhance-raisr-sharp-images-with-machine.html
[18:10:35 CEST] <cslcm> https://www.youtube.com/watch?v=dVU_sQGxyMg
[18:11:09 CEST] <cslcm> sorry, this one shows the performance better https://www.youtube.com/watch?v=n0S7CLfXELQ
[18:36:16 CEST] <adrian_1908> Can I assume that most ffmpeg installation come with ffprobe?
[18:36:24 CEST] <adrian_1908> s/installation/installations
[18:42:48 CEST] <durandal_1707> it should
[18:43:04 CEST] <cslcm> though you can get the same info out of ffmpeg itself
[18:44:22 CEST] <durandal_1707> not really, you can get varios outputs formatted in many ways
[18:45:03 CEST] <adrian_1908> cslcm: with the same formatting? I saw that ffprobe can print audiobook metadata (chapters) in nice JSON format.
[18:48:23 CEST] <adrian_1908> A related question: Can I pass a time slice (from, to) in milliseconds to ffmpeg? It seems the input format is human-readable only.
[18:54:56 CEST] <adrian_1908> Hmm, I think I can just add a colon and it'll work, i.e. split ms into `s.ms`. Will try later, bye now! :)
[19:13:26 CEST] <Lite89> Hello
[19:41:27 CEST] <Lite89> I'd wish to ask, about way(s) of concatenation of encrypted transport stream (.TS) segments via FFMPEG.
[19:42:59 CEST] <Lite89> As usual, I am downloading .TS segments to some local path.
[19:44:03 CEST] <Lite89> Than creating 'MyList.TXT' file where placing string by string paths to local files needed to be concantenated.
[19:45:11 CEST] <Lite89> In format: file 'C:\Path_to_local_file.TS'
[19:47:51 CEST] <Lite89> When segments are enrcypted, there https://ffmpeg.org/ffmpeg-protocols.html at paragraph 3.5 'crypto' I see that provided format should be exchanged to
[19:48:20 CEST] <Lite89> file 'crypto+C:\Path_to_local_file.TS' or file 'crypto:C:\Path_to_local_file.TS'
[19:48:44 CEST] <Lite89> BUT I can not understand, how to set key and iv parameters.
[19:49:43 CEST] <Lite89> Even where it should be set up: in MyList txt file or at command string to concatenate via FFMPEG.
[19:49:53 CEST] <Lite89> Anybody?
[19:53:04 CEST] <Lite89> I tried command like FFMPEG -f concat -safe 0 -i 'C:\FFMPEG\MyList.TXT' -key 'value_in_HEX' -iv 'value_in_HEX' -c copy C:\output_concatenated_file.MP4 but it sends me an error:
[19:53:24 CEST] <Lite89> decryption key not set
[19:53:44 CEST] <Lite89> So how to set this correctly??
[19:54:59 CEST] <Lite89> Anybody? Please!
[19:55:51 CEST] <ChocolateArmpits> Lite89, those parameters are meant to be specified as input options
[19:56:09 CEST] <ChocolateArmpits> I'm not sure if they will work with concat list right away
[19:56:31 CEST] <ChocolateArmpits> You may want to decode each ts file first, then concatenate them
[19:56:46 CEST] <ChocolateArmpits> Lite89, what do you think?
[19:56:52 CEST] <Lite89> so what are they for? where else it could be used 'crypto+URL' command?
[19:57:32 CEST] <ChocolateArmpits> instead of doing concatenation right away, decrypt each separately ffmpeg -key key -iv iv -i 'crypto+url' output.ts
[19:57:42 CEST] <Lite89> Ah, good advise, sir ChocolateArmpits
[19:58:04 CEST] <Lite89> sec
[19:58:29 CEST] <ChocolateArmpits> of course don't forget -c copy
[20:26:00 CEST] <Lite89> Sir @ChocolateArmpits! :D
[20:26:12 CEST] <ChocolateArmpits> Lite89, it works?
[20:26:48 CEST] <Lite89> Yep. What you seggested works. But pretty strange in compare with how it decoding by OpenSSL.
[20:28:30 CEST] <Lite89> I mean, size of files and some other parameters witch are results of decrypting via OpenSSL and via FFMPEG are totally different
[20:31:09 CEST] <Lite89> Does it mean that FFMPEG not just clear decodes encrypted file? It also converts it deleting options it doesnt understand?
[20:32:10 CEST] <ChocolateArmpits> well it's both rewrapping it
[20:32:24 CEST] <ChocolateArmpits> is the stream fully encoded, or only the content?
[20:34:06 CEST] <Lite89> ye, but size of decoded files which are result of OpenSSL activity is same or almost same as originally encoded files
[20:36:30 CEST] <Lite89> ATM for example as soon as result of OpenSSL activity size is 247408 bytes (encrypted source file is 247424 bytes), result of FFMPEG size is 196084 bytes.
[20:37:08 CEST] <Lite89> I believe its not normal and OpenSSL saves much more original information after decrypting than FFMPEG :C
[20:37:59 CEST] <Lite89> I meant that
[20:38:19 CEST] <ChocolateArmpits> welp
[20:38:43 CEST] <ChocolateArmpits> you can try decrypting with that and then concatenate using cat :3
[20:39:04 CEST] <Lite89> Yes, and here I met base issue...
[20:39:13 CEST] <Lite89> I will provide it via screenshot
[20:41:08 CEST] <Lite89> I tried to download some AES-128 encrypted content via FFMPEG. But having TLS-issue restricting me to ask FFMPEG to proceed any HTTPS URLs, even this is just key URL
[20:41:45 CEST] <Lite89> So I downloaded TS segments and decrypted them to concatenate
[20:42:49 CEST] <Lite89> When I started concatenation, I started to see many errors about each of 400+segments
[20:43:57 CEST] <Lite89> ATM, I tried to download what I need via another PC where it's everything OK with TLS, and did not meat any issue.
[20:44:07 CEST] <Lite89> meet*
[20:45:46 CEST] <Lite89> so I believe thats because decrypted by OpenSSL segments contains some extract information which FFMPEG doesn't understand
[20:46:59 CEST] <Lite89> and if FFMPEG was rewrapping it, it just created totaly different decoded segments which does not contain such information
[20:51:24 CEST] <Lite89> Well, look:
[20:52:31 CEST] <Lite89> when I'm just trying to download it via FFMPEG, providing external URL to M3U list, it starts to download it with just these list of errors:
[20:52:35 CEST] <Lite89> 1) non-existing SPS 0 referenced in buffering period
[20:52:46 CEST] <Lite89> 2) SPS unavailable in decode_picture_timing
[20:52:56 CEST] <Lite89> 3) non-existing SPS 0 referenced in buffering period
[20:53:07 CEST] <Lite89> 4) SPS unavailable in decode_picture_timing
[20:53:23 CEST] <Lite89> than it's everything going fine.
[20:53:59 CEST] <Lite89> I mean it doesn't shows any errors during downloading process
[20:55:27 CEST] <Lite89> When I'm rying concatenation of segments, decrypted by openSSL, I meet this list of errors (for each segment):
[20:58:28 CEST] <Lite89> SPS unavailable in decode_picture_timing, non-existing PPS 0 referenced, decode_slice_header error, no frame!, non-existing SPS 0 referenced in buffering period
[20:58:38 CEST] <Lite89> Why?
[20:58:56 CEST] <Lite89> Any ideas?
[21:02:24 CEST] <Lite89> somethhing like that https://stackoverflow.com/questions/42432898/live-tv-recording-ts-to-mp4-with-ffmpeg. so you don't need my screenshot.
[21:05:40 CEST] <Lite89> Sir @ChocolateArmpits :D
[21:08:09 CEST] <ChocolateArmpits> Lite89, are you sure the concatenated segments are in correct order?
[21:08:52 CEST] <ChocolateArmpits> alternatively you can try concatenating the segments using cat. simple byte concatenation should work with ts provided the files are in correct order
[21:39:25 CEST] <Lite89> definately sure.
[21:41:52 CEST] <Lite89> in already concatenated files it looks everything ok, maybe except buffering.
[21:44:10 CEST] <Lite89> cat. simple byte concatenation? Can you explain it in detail?
[21:44:21 CEST] <Lite89> please: D thanks a lot :D
[21:47:59 CEST] <ChocolateArmpits> Lite89, are you on Linux ? https://linux.die.net/man/1/cat
[21:48:29 CEST] <ChocolateArmpits> so: cat file1.ts file2.ts > final.ts
[23:15:58 CEST] <Lite89> @ChocolateArmpits, nope, Windows. You mean without -i 'MyList.TXT' but proving local paths to segments directly in command string?
[23:16:15 CEST] <ChocolateArmpits> Lite89, oh then you can use copy /b
[23:16:44 CEST] <ChocolateArmpits> I think it was something like copy /b file1.ts + file2.ts final.ts
[23:17:13 CEST] <Lite89> for this list of segments even -c copy command is worst
[23:17:46 CEST] <Lite89> only -c:a copy -c:v copy or -vcodec copy -acodec copy
[23:18:57 CEST] <Lite89> because there is integrated 2 subtitle streams integrated as base64 data in media content itself. And I need to save it as in source
[23:19:36 CEST] <Lite89> -c copy erasing it and saving traditionally only video and audio streams
[00:00:00 CEST] --- Mon Sep 24 2018