[Ffmpeg-devel-irc] ffmpeg.log.20170701

[burek]

[01:07:26 CEST] <Hopper_> Hey all, anyone have a moment to troubleshoot my "Unable to find a suitable output format for [out]"?
[01:09:36 CEST] <DHE> you probably have a syntax error in your commandline
[01:10:01 CEST] <DHE> I assume [out] is a filter_complex pipeline element
[01:10:09 CEST] <Hopper_> I'm SURE I do, just started on ffmpeg this morning.
[01:10:26 CEST] <DHE> pastebin plz
[01:11:01 CEST] <Hopper_> working on it, the device which is running the command is not wan connected.
[01:12:28 CEST] <Hopper_> https://pastebin.com/e7Fuu0Gd
[01:13:37 CEST] <DHE> after -filter:v you should have a space
[01:14:08 CEST] <DHE> you've accidentally glued two parameters into one
[01:14:35 CEST] <Hopper_> Good catch, thanks.
[01:15:27 CEST] <Hopper_> Now it says "output with label 'out' does not exist in any defined filter graph, or was already used elsewhere."
[01:16:14 CEST] <DHE> ah, yes. -filter:v should be -filter_complex
[01:17:23 CEST] <Hopper_> Look at you!  That's got it working now.
[01:17:36 CEST] <Hopper_> Where do I put -threads 0 in here?
[01:18:08 CEST] <DHE> that's usually the default. otherwise, it goes just before the output file typically
[01:19:12 CEST] <Hopper_> Hm, I'm getting a heap of brown text telling me "Past duration 0.0numbers too large"
[01:21:26 CEST] <DHE> test the output to see if it looks okay. it might not be a big deal, but make sure
[01:22:07 CEST] <Hopper_> no video on the output, and vlc doesn't even show a timecode.
[01:22:20 CEST] <Hopper_> But the file is growing in size.
[01:23:24 CEST] <DHE> hmm... I suspect mepg-ts doesn't support mjpeg
[01:23:45 CEST] <Hopper_> It works if it's a single data stream.
[01:24:42 CEST] <Hopper_> DHE: I appreciate your help, but my work day is over.  I'll be back on wed after the holiday, I hope you'll be around then.  You've already been very helpful!
[01:32:34 CEST] <DHE> plenty of other people here. I'm just at my keybaord and happen to have an idea what it might be
[03:11:56 CEST] <kepstin> alexpigment: so, my industry friends got back to me - and, well, the answer is "it depends on the tv", some will only support 24+50 or 24+60 with different models per territory, others will support multiple frequencies. so, yeah :/
[07:38:38 CEST] <memphisw>  /script load usercount.pl
[10:44:55 CEST] <hiihiii> hello
[10:45:36 CEST] <hiihiii> how do I use the ocr filter? I can't seem to find a way to output results to a file
[10:46:19 CEST] <durandal_1707> hiihiii: with metadata filter
[10:46:59 CEST] <durandal_1707> -vf ocr,metadata=print
[10:48:43 CEST] <hiihiii> yes in the doc it says something about that filter "The filter exports recognized text as the frame metadata" but I didn't understand well
[10:49:41 CEST] <durandal_1707> hiihiii: use command i gave you
[10:50:12 CEST] <hiihiii> so it's -vf ocr,metadata=print:file=result.txt
[10:50:20 CEST] <hiihiii> right?
[10:50:25 CEST] <durandal_1707> yes
[10:50:31 CEST] <hiihiii> ok thx
[10:51:59 CEST] <hiihiii> how accurate is the recognition? my input is something like this "0000015    F2   0"
[10:52:15 CEST] <hiihiii> each frame is just one line
[10:52:48 CEST] <durandal_1707> hiihiii: depends how on many factors
[10:54:29 CEST] <hiihiii> ocr's whitelist is the letters that I only want to catch
[10:56:01 CEST] <hiihiii> how do I set that? -vf ocr=whitelist='0123456789ABCDEF',metadata=print:file=result.txt
[10:57:43 CEST] <durandal_1707> yes, read docs
[11:06:24 CEST] <thebombzen> hiihiii: remember that it uses tesseract for OCR so you need to have the appropriate tesseract language packs installed. I think English is bundled in with the windows zeranoe builds. on Linux make sure you install tesseract-data-eng or equivalent
[11:08:19 CEST] <hiihiii> isn't English the default
[11:13:46 CEST] <thebombzen> well on Arch, for example, it doesn't come packaged in with tesseract and you need to specifically install tesseract-data-eng
[11:13:59 CEST] <thebombzen> on other systems, English might come bundled in
[11:46:35 CEST] <hiihiii> ummm is there a windows build somewhere with tesseract?
[18:14:13 CEST] <thoraxe> this may be a dumb question but i am playing with an rstp device that seems to have a borked audio stream. I don't really care about the audio, I'd be happy just to drop it
[18:14:34 CEST] <thoraxe> trying to figure out the magic google search for dropping/ignoring audio in a stream
[18:14:40 CEST] <DHE> -an
[18:15:46 CEST] <thoraxe> ah, perfect
[18:15:54 CEST] <thoraxe> now to figure out why youtube live doesn't think i'm actually sending it data
[18:19:14 CEST] <ChocolateArmpits> Is there any way to control image compression for BMP files?
[18:19:50 CEST] <kerio> bmp is not compressed
[18:20:27 CEST] <kerio> oh wait, apparently it can be :o
[18:21:14 CEST] <BtbN> it has some very basic compression
[18:21:17 CEST] <BtbN> but why not just use png?
[18:21:54 CEST] <ChocolateArmpits> bitmap stores image data as a simple array if there's no compression, png has filter information on every line start
[18:22:48 CEST] <DHE> "ffmpeg -h encoder=bmp" does not indicate any options supported
[18:24:30 CEST] <ChocolateArmpits> png makes use of compression_level argument to specify compression level, that however doesn't work for bmps
[18:25:20 CEST] <BtbN> But why the need for bmp?
[18:25:29 CEST] <BtbN> png is also lossless, but compresses way better
[18:26:50 CEST] <ChocolateArmpits> I want to pack opus files into images and then read those using -skip_initial_bytes. As I said before, png has filter information on every line that kind of messes up playback somewhat
[18:27:29 CEST] <thoraxe> well, i got twitch working
[18:27:40 CEST] <thoraxe> youtube live doesn't think i'm streaming to it after a few moments
[18:29:13 CEST] <BtbN> but why? oO
[18:29:19 CEST] <BtbN> @opus in bmp
[18:30:08 CEST] <ChocolateArmpits> cause there are places where anything but only a select few image formats are allowed to be posted
[18:30:24 CEST] <ChocolateArmpits> and opus is really efficient for audio exchange
[18:31:40 CEST] <BtbN> So you want to upload audio to an image board of some kind? It won't play anywhere though?
[18:32:23 CEST] <ChocolateArmpits> well that's why using a backwards command line to seek to a specific part of the file will
[18:32:46 CEST] <ChocolateArmpits> something like ffplay -skip_initial_bytes 46 -f ogg image.png
[18:33:02 CEST] <ChocolateArmpits> actually that's how I do it atm with pngs
[18:33:25 CEST] <ChocolateArmpits> though I'm not happy with filter info
[18:34:17 CEST] <ChocolateArmpits> uncompressed pngs have each line start with 0x00 inserted, so on playback it's corrupted for a few samples on every line
[18:34:51 CEST] <ChocolateArmpits> I can increase width, bit depth and channel count to offset the problem ,but it won't remove it completely
[18:37:12 CEST] <kerio> can't you just append a rar to a jpeg
[18:37:17 CEST] <kerio> like in the good ol times
[19:57:26 CEST] <ChocolateArmpits> aww yes finally working. Bmps actually aren't compressed, I just didn't know they are written bottom up. Vflip solves this  however :).
[20:03:10 CEST] <furq> you could just use tif
[20:06:37 CEST] <alexpigment> Hey guys, I'm making a music video playlist for a party we're throwing tonight. I've got a huge collection of DVD rips and stuff, but of course the volumes are all over the place. Does anyone have any advice on filters that would give a perceptually similar level of volume?
[20:07:05 CEST] <alexpigment> I'm thinking about just converting everything to MKV, copying the video stream, and filtering the audio into PCM
[20:08:31 CEST] <furq> you could use replaygain if your player supports it
[20:08:49 CEST] <furq> i'm pretty sure mpv supports it now
[20:11:26 CEST] <JEEB> ChocolateArmpits: the height can be negative as well as positive in BMPs
[20:11:34 CEST] <JEEB> read the header and you'll know how to read them
[20:16:23 CEST] <alexpigment> furq: replaygain would just be a simple normalizer, right?
[20:17:06 CEST] <furq> it's not even that, it's just player auto-amplification to -6dB
[20:17:41 CEST] <alexpigment> but there's no dynamic range compression to make the music sound like it's at the same level, right?
[20:17:47 CEST] <furq> no
[20:17:52 CEST] <furq> it's pretty much zero effort though
[20:17:56 CEST] <alexpigment> Kodi supports it, which is what I'll be using, but I'm thinking I need to use compression
[20:17:56 CEST] <furq> fb2k will do it
[20:18:21 CEST] <alexpigment> does fb2k play videos yet?
[20:18:27 CEST] <alexpigment> i thought it was just audio
[20:18:35 CEST] <furq> i mean it'll add the replaygain tags to mkv
[20:18:46 CEST] <alexpigment> gotcha
[20:19:27 CEST] <furq> i'd have thought it would work well for music videos
[20:20:17 CEST] <alexpigment> well, the problem is just that music is mixed very differently, so sometimes the snare drum is loud, and if you normalize, it'll base it on the peaks, but the average of the peaks and the valleys is what contributes to perceived volume
[20:20:32 CEST] <furq> oh nvm i'm wrong it's not just peak normalisation
[20:20:48 CEST] <alexpigment> oh cool
[20:21:19 CEST] <alexpigment> (i probably could have just googled it when you suggested it :) )
[20:21:21 CEST] <alexpigment> sorry about that
[20:21:21 CEST] <furq> it does just apply a straight amplitude gain though
[20:21:24 CEST] <alexpigment> looking into it
[20:21:55 CEST] <alexpigment> well, as long as it does it by doing a scientific analysis of the original perceived volume, it's not a huge deal
[20:22:12 CEST] <alexpigment> i've got lots of headroom on my speakers, so volume spikes at the peaks are zero problem
[20:22:23 CEST] <ChocolateArmpits> you could look into ebur128 filters if you have time
[20:22:42 CEST] <furq> isn't that the same thing
[20:22:49 CEST] <ChocolateArmpits> no
[20:23:49 CEST] <alexpigment> I'll check out ebur128 in a bit
[20:24:04 CEST] <alexpigment> furq: you mentioned adding replaygain to MKV files with foobar2k
[20:24:21 CEST] <alexpigment> is it pretty straightforward to do that or are there any tricks I should know about
[20:24:28 CEST] <furq> not really
[20:25:05 CEST] <furq> your only real choice is "scan per-file track gain" or "scan selection as single album"
[20:25:17 CEST] <furq> maybe the second one would be better if these are part of a compilation
[20:26:00 CEST] <alexpigment> nah, it's a completely hand-picked video playlist
[20:26:16 CEST] <alexpigment> around a hundred from my collection, so it's not going to be similar in volume
[20:26:30 CEST] <alexpigment> anyway, i'll give it a shot. thanks for the heads up
[22:25:33 CEST] <nicolas17> I managed to decode 1 (one) frame from a video \o/
[22:25:39 CEST] <JEEB> 'grats
[22:25:41 CEST] <nicolas17> (with the API)
[22:25:43 CEST] <JEEB> that's the first thing
[22:25:51 CEST] <JEEB> (I did that in my matroska thumbnailer for windows)
[22:25:51 CEST] <nicolas17> and output it to a ppm
[22:26:03 CEST] <JEEB> in my case the output was a BMP structure because windows wanted one)
[22:26:18 CEST] <JEEB> I also implemented custom IO since Windows would be giving the data through their IStream API
[22:26:36 CEST] <nicolas17> hm the result is wrong
[22:26:55 CEST] <nicolas17> looks like I'm getting all the red samples, then all green, then all blue
[22:27:01 CEST] <nicolas17> while ppm expects r,g,b,r,g,b,r,g,b
[22:28:08 CEST] <JEEB> sounds like you're either taking data as decoded or so
[22:28:37 CEST] <JEEB> you should know what your decoded pictures are in (which PIX_FMT)
[22:28:47 CEST] <nicolas17> well I'm getting YUV, so I'm converting it to AV_PIX_FMT_RGB24 with swscale
[22:28:49 CEST] <JEEB> and then if you need to convert you can utilize various functionality)
[22:28:56 CEST] <JEEB> oh, that should be not planar
[22:29:02 CEST] <JEEB> unless you use the planar color space
[22:29:08 CEST] <JEEB> (there's a specific planar RGB thing)
[22:29:19 CEST] <JEEB> then you're handling the data after the conversion incorrectly I think?
[22:29:37 CEST] <JEEB> check the pix_fmts on all places that they are what you expect
[22:30:15 CEST] <nicolas17> okay I'm an idiot, I'm decoding into 'frame', using swscale to convert into another AVFrame called 'frameRGB', then passing 'frame' to my save_ppm function \o/
[22:30:23 CEST] <JEEB> :D
[22:30:35 CEST] <nicolas17> so I guess this is the planar YUV I'm looking at? or whatever it is
[22:30:40 CEST] <JEEB> yes
[22:30:50 CEST] <JEEB> if the decoded image is planar YCbCr then that's what it is
[22:32:19 CEST] <nicolas17> can I convert an AVPixelFormat to a string name?
[22:32:29 CEST] <JEEB> yea
[22:32:35 CEST] <JEEB> there was a utility function for it
[22:32:39 CEST] <JEEB> which the command line apps are also using
[22:32:41 CEST] <nicolas17> I'm digging through doxygen
[22:33:59 CEST] <JEEB> av_get_pix_fmt_string
[22:34:01 CEST] <JEEB> I think
[22:34:15 CEST] <JEEB> in libavutil
[22:35:28 CEST] <nicolas17> yeah so input is yuvj422p
[22:37:01 CEST] <JEEB> so full range 4:2:2 YCbCr
[22:40:38 CEST] <nicolas17> okay, I got a black image and a segfault, turned out sws_scale wasn't succeeding
[22:40:50 CEST] <nicolas17> now it did (wrong arg) and I have a valid image \o/
[22:41:12 CEST] <nicolas17> and valgrind says I'm freeing everything I should
[22:42:01 CEST] <nicolas17> it's a hundred lines of horror though
[22:57:13 CEST] <nicolas17> JEEB: I'm decoding JPEGs with image2, and I need to access their EXIF data, does ffmpeg decode EXIF?
[22:57:36 CEST] <nicolas17> otherwise, can I access the filename of the current frame so I can feed it into an EXIF library?
[22:58:08 CEST] <JEEB> no idea about EXIF
[22:58:20 CEST] <JEEB> you can check the libavformat/libavcodec demuxer related
[22:58:23 CEST] <JEEB> (or decoder)
[22:59:04 CEST] <durandal_170> iirc its available as frame metadata
[22:59:25 CEST] <nicolas17> do I need to request that first?
[22:59:30 CEST] <nicolas17> frame->metadata is NULL
[23:00:01 CEST] <durandal_170> wasnt it tiff?
[23:01:51 CEST] <nicolas17> nvm I'm an idiot again, I was looking at *frame with a debugger but I hadn't stepped past avcodec_receive_frame yet
[23:03:42 CEST] <nicolas17> frame->metadata is an AVDictionary with all the EXIF tags :)
[23:04:14 CEST] <durandal_170> \Ë/
[23:13:19 CEST] <nicolas17> why does uint8_t** -> const uint8_t* const* conversion give a warning? it's totally safe :/
[23:22:24 CEST] <thebombzen> nicolas17: Because if you cast char** to const char* const*, then any caller receiving the const char* const* pointer will assume that it's const
[23:23:20 CEST] <thebombzen> suppose I have: char **x; inside a() and I cast it to (const char *const*) and pass that casted variable to b()
[23:23:50 CEST] <rcombs> what's wrong with that
[23:23:51 CEST] <durandal_170> too many indirections
[23:23:56 CEST] <nicolas17> that doesn't let b do something it shouldn't
[23:24:02 CEST] <thebombzen> well b() is going to assume that the pointer passed is const. However, a() could still modify it.
[23:24:20 CEST] <thebombzen> b() isn't going to do something it shouldn't, but from b()'s perspective, a() can do something it shouldn't.
[23:24:22 CEST] <rcombs> adding const never creates opportunities for invalid accesses
[23:24:28 CEST] <rcombs> _removing_ const can
[23:24:36 CEST] <nicolas17> thebombzen: that's true for single indirection too, if I have a char[42
[23:24:39 CEST] <nicolas17> argh keyboard
[23:24:57 CEST] <thebombzen> It never creates opportunities, sure, but it creates extra assumptions or assurances that it really is const.
[23:25:00 CEST] <nicolas17> thebombzen: that's true for single indirection too, if I have a char[42] and I pass it as a const char* to b(), I can still modify it, but I'm pretty sure that doesn't warn
[23:25:15 CEST] <thebombzen> Pretty sure it does
[23:25:39 CEST] <rcombs> it does not
[23:25:41 CEST] <nicolas17> so passing a char[42] to strlen warns?
[23:26:57 CEST] <kerio> nicolas17: http://c-faq.com/ansi/constmismatch.html actual reason
[23:27:52 CEST] <thebombzen> so it seems, it does not warn.
[23:27:57 CEST] <thebombzen> Goes to show how much I know about C.
[23:31:10 CEST] <nicolas17> kerio: thanks, that explains the hole it would open
[23:38:10 CEST] <marcurling> Hello, is there an auto-crop function to get rid/cut black stripes from letterbox films?
[23:40:16 CEST] <kerio> cropdetect
[23:48:05 CEST] <marcurling> ty kerio. Which tool would you recommend to set the (correct) Display aspect Ration to an anamorphosed movie?
[00:00:00 CEST] --- Sun Jul  2 2017