[Ffmpeg-devel-irc] ffmpeg.log.20140607

burek burek021 at gmail.com
Sun Jun 8 02:05:01 CEST 2014


[02:54] <BtbN> "No accelerated colorspace conversion found from yuv420p to rgba." is that a problem with my self built version of libswscale, or is there simply no accelerated way to do that?
[05:14] <deddisk> im having trouble recording my audio output from alsa when i screenrecord with x11grab; if I use -f alsa -i hw:0 or -f alsa -i default, I record my mic, but I'm not sure how to record the audio out instead
[05:15] <c_14> afaik alsa can record from any device listed in `arecord -l'
[05:15] <deddisk> arecord -l only shows hw:0 tho :/ and it is listed as "analog"
[05:16] <c_14> ye, if you want to record audio output you'll probably have to use the loopback device
[05:35] <benlieb> I have a video (dvd rip) that is in the following format:  720x362 [SAR 8:9 DAR 320:181
[05:35] <benlieb> It will now be living a digital life and be subject to various manipulations, crops and resizings. Should I convert this to square pixels?
[05:35] <benlieb> That seems to be the digital way to do things.
[05:36] <benlieb> c_14: ^
[05:40] <c_14> It's really up to you. If you're going to be doing things like cropping and resizing, square pixels are easier to work with because they require less math. I also personally prefer square pixels, but that might just be me.
[05:41] <benlieb> c_14: what filter would I apply to change the SAR and specify the size?
[05:42] <benlieb> c_14: ^
[05:43] <c_14> -vf scale=iw*sar:ih should do the trick
[05:43] <c_14> you might also need setsar=1
[05:48] <benlieb> c_14: there seem to be actual setdar, setsar filters. Are they a viable option?
[05:49] <devsherif> I cannot build ffmpeg for android using Mac OSx, someone can help please ?
[05:49] <c_14> benlieb: I'm not sure if the setsar filter actually modifies the resolution and pixels or just forces the sar to 1 thereby messing with the dar.
[05:51] <benlieb> c_14: how would the latter be a good thing? it would basically just get the video to report it's SAR incorrectly?
[05:52] <c_14> ye, which is why I said I'm not sure if the setsar filter does what you want
[05:54] <benlieb> c_14: I'm not sure why anyone would want that, so I'm surprised it would do that. Unless you see a benefit to that.
[05:54] <benlieb> I'll try the scale first.
[05:54] <benlieb> Just as soon as I understand it :)
[05:55] <c_14> I'm guessing the use would be if a video has an incorrect sar set or something, but I could be wrong.
[05:57] <benlieb> How exactly does this change to square pixels? -vf scale=iw*sar:ih
[05:57] <benlieb> c_14: ^
[06:00] <c_14> The sar is the ratio that says how large each 'pixel' is, multiplying the input width by the sar chops each of the n rectangular 'pixels' into m square 'pixels'
[06:03] <benlieb> I guess I just don't understand what it getting set. Your setting scale=a_new_ratio
[06:04] <benlieb> wouldn't that just change the dimensions?
[06:04] <c_14> the dar stays the same, the sar changes
[06:04] <benlieb> I"m confused, and I just read the wikipedia entry on PAR/SAR/DAR
[06:04] <benlieb> ug
[06:05] <c_14> dar = (width/height)*sar
[06:05] <c_14> the scale filter is just multiplying the sar into the (width/height) and setting a new sar of 1
[06:06] <c_14> so dar = (width*sar/height)
[06:06] <c_14> scale=iw*sar:ih
[06:07] <benlieb> i'm missing something fundamental. the SAR is just some metavariable that says how to display, or is actually a result of encoding of some kind?
[06:07] <c_14> it's basically just a metavariable
[06:07] <benlieb> So ffprobe says this 720x362 [SAR 8:9 DAR 320:181], and Mac finder says 640×362
[06:07] <benlieb> I'm so confused.
[06:08] <benlieb> Why doesn't my video just have ONE size.
[06:08] <c_14> 720*(8/9) = ?
[06:08] <c_14> 640
[06:08] <benlieb> I know math, I just don't understand how this relates to if a pixel is square or not
[06:08] <benlieb> how is an unsquare pixel displayed
[06:09] <benlieb> is it "faked"  if the SAR is not 1:1?
[06:09] <c_14> An rectangular pixel uses numerous square pixels.
[06:09] <c_14> basically
[06:09] <benlieb> right
[06:09] <benlieb> so by resetting the SAR, it won't do that, and then won't the video be warped?
[06:10] <c_14> so what the scale filter is doing is the same thing that every video player does, just more permanently
[06:10] <c_14> benlieb: if you just set the sar, yes
[06:10] <c_14> That's why I recommended the scale.
[06:10] <benlieb> Or would resetting the SAR transcode the video so that the new SAR actually works
[06:10] <c_14> That depends on the setsar filter, I'm not entirely sure about the inner workings thereof. But if you were to manually change the sar of the video, the video would be warped.
[06:12] <benlieb> c_14 ok that's what I thought
[06:13] <benlieb> another things is that wikipedia is calling SAR storage aspect ration, while ffmpeg docs are calling it sample aspect ration
[06:13] <benlieb> ratio
[06:14] <c_14> the sar is the physical aspect ratio of the file, if you were to count the pixels you would get the sar. Not sure if it matters how you name it.
[06:15] <benlieb> the docs are unclear as to what actual option is getting set when you do scale=x:y
[06:15] <benlieb> what is that short for?
[06:15] <benlieb> scale=sar=x:y ?
[06:16] <c_14> scale messes around with the number of pixels
[06:18] <benlieb> Bear with me here. I just found the docs say that scale=w=200:h=100 This is equivalent to: scale=200:100
[06:19] <benlieb> But I don't see how this relates to square or nonsquare pixels.
[06:19] <benlieb> You could have a frame images of 200:100 at square or not square pixels
[06:20] <benlieb> just changing the dimensions doesn't seem like it wouldn't do anything in particular to pixel type
[06:20] <c_14> Ok, back to the beginning. I'm going to use your video as an example.
[06:21] <c_14> The video has 340 x 362 pixels.
[06:21] <benlieb> ok :)
[06:21] <c_14> These pixels are anamorphic, ie not square.
[06:21] <benlieb> where did you get that?
[06:21] <c_14> The sar is 8:9
[06:21] <c_14> Square pixels have a sar of 1:1
[06:21] <benlieb> c_14 so you derived it?
[06:22] <benlieb> ok
[06:22] <c_14> A sar of 8:9 means that each pixel is 8 units wide and 9 units high
[06:22] <c_14> Therefore, when you play the video it is displayed as having 720 x 362 pixels.
[06:23] <c_14> So your video has 640 x 362 pixels but is displayed as having 720 x 362 pixels.
[06:23] <c_14> With me so far?
[06:23] <benlieb> still thinking, one sec...
[06:24] <c_14> Wait, I messed up. everywhere I said 720 I meant to say 640 and vice verse.
[06:24] <c_14> The rest still stands though.
[06:25] <c_14> And that 340 should be a 720
[06:25] <benlieb> aaha
[06:25] <benlieb> ok can we start over :)
[06:25] <c_14> sure
[06:25] <c_14> Your video has 720 x 362 pixels.
[06:25] <c_14> The sar is 8:9, therefore the video is displayed as having 640 x 362 pixels.
[06:26] <benlieb> ok THAT makes sense
[06:26] <c_14> Now all the scale filter is doing, is taking the video that has 720 x 362 pixels and converting it to a video that has 640 x 362 pixels.
[06:27] <benlieb> what makes most sense to me to do this would be to say something like scale=sar=1:1
[06:28] <c_14> I'm pretty sure scale doesn't have that option.
[06:28] <benlieb> I guess I'm thinking that scale would be used for actually intentionally warping your video, if need be. But I guess that you could never do that with scale?
[06:28] <c_14> You can, just not with the command I gave you.
[06:29] <benlieb> this is what I don't understand
[06:29] <c_14> scale takes the input video and converts it to the width and height you give it.
[06:29] <c_14> I don't want to change the height so I just use ih (the input height) for that part.
[06:30] <c_14> I do however need to change the width.
[06:30] <benlieb> v_14: why wouldn't it convert it to the new w and h without changing the pixel type?
[06:31] <c_14> I'm pretty sure the scale filter uses square pixels (a sar of 1:1) as standard but I'm not sure which is why I said you might need to append the setsar=1 filter after that one.
[06:31] <benlieb> I guess what I don't understand is why does changing the w and h imply a pixel type change
[06:31] <c_14> It doesn't.
[06:31] <c_14> The pixel type only changes when you change the sar.
[06:32] <c_14> Thing is I'm pretty sure the scale filter resets the sar to 1:1 by default.
[06:32] <benlieb> c_14 ok so I would need to change the scale and then use the setsar command
[06:32] <benlieb> THAT makes sense.
[06:32] <c_14> That would be the foolproof method.
[06:33] <benlieb> ok I'll play around with that. I guess that answers my own earlier question about why someone would want to set the SAR
[06:33] <benlieb> tnx for your -i
[06:34] <benlieb> lol
[06:34] <benlieb> I owe you a coupla drinks
[06:35] <c_14> no problem
[06:56] <benlieb> so the command works setting the w x h, and doesn't affect the DAR
[06:57] <benlieb> the new video is now fully half the size (in MB) of the orig. Does that translate into quality loss?
[06:57] <c_14> What command did you use to encode? Ie what encoder with what quality settings?
[06:58] <benlieb> c_14: I just did this ffmpeg -i 1.mp4 -vf scale=iw*sar:ih 1_scaled.mp4
[06:59] <c_14> Then yes, you probably had quality loss there. I'm not sure what the default quality settings ffmpeg uses are, but they aren't much on the quality side.
[07:00] <benlieb> I'll look into that.
[07:00] <benlieb> but my real task is to try to get this vid into a standard shape so that I can concat things to it.
[07:00] <benlieb> It's currently  640x362 [SAR 1:1 DAR 320:181]
[07:01] <benlieb> I'd like this to be 640x360 [sar 1:1 dar 16:9]
[07:01] <benlieb> I"m not sure where those extra pixels came from, probably the original dvd
[07:01] <benlieb> would I do a crop and then a setdar?
[07:01] <c_14> Just a crop should be fine.
[07:25] <Raven_> hi all
[07:29] <Raven_> can anyone point me to the right direction for finding rtmpdump v2.5 source. Im trying to build a Linux media center and it needs the RTMPE 10 handshake for crunchyroll. much appreciated
[07:43] <sacarasc> Raven_: Unless git clone git://git.ffmpeg.org/rtmpdump
[07:43] <sacarasc>  is up to date, I haven't a clue. And if it's not, then the project has cut its ties with ffmpeg.
[07:50] <Raven_> sacarasc ty for the reply, unfortunately rtmpdump 2.5(unofficial) was released in 2012 by Xeebo. It had a issue with adobe since they made a working flash handshake.
[07:50] <JEEBsv> there was a version for a while that used a binary dump or something from Flash player, which is why it was taken down IIRC. Also AFAIK it still isn't really needed for Crunchy
[07:50] <Raven_> it is hard to find
[07:51] <JEEBsv> you just need a new enough implementation in general :P
[07:51] <JEEBsv> (I might be incorrect, but as far as I can remember...)
[08:09] <Raven_> JEEBsv ty, unfortently im trying to get RTMPE streaming to work. RTMPE is adobe DRM streaming protocol
[08:09] <Raven_> here is a good info link on it https://en.wikipedia.org/wiki/Protected_Streaming
[08:29] <Raven_> Also fyi here is what im trying to do https://code.google.com/p/tv-maxe/
[08:58] <Raven_> im heading out, going to try tomorrow. if this not work then going to XBMC. Have a good night
[12:21] <rsdrsdrsd> is there any difference betwee setsar=1 and setsar=sar=1
[12:21] <rsdrsdrsd> and if so, what is the difference?
[12:22] <ubitux> the difference is in the case sar option is moved
[12:23] <ubitux> which shouldn't happen (we add them at the end to avoid breaking the short usages)
[12:24] <ubitux> rsdrsdrsd: http://ffmpeg.org/ffmpeg-filters.html#Filtergraph-syntax-1 look for "arguments is a string ..."
[14:09] <Ofek> hello! is there a way to insert a small ac3 in the middle of an existing one?
[14:10] <Ofek> or is there a command to split audio (ac3) into 2 parts at certain time?
[14:10] <c_14> look at the asplit filter in combination with a filter complex and the concat filter
[14:35] <DeadSix27> can ffmpeg handle pgs/.sup files in any form?
[14:35] <DeadSix27> as i get -> Title2.sup: Invalid data found when processing input
[15:02] <daedeloth> I need to combine 4 videos into one video (sortof like security cam footage)
[15:03] <daedeloth> not live
[15:03] <daedeloth> is that hard to do?
[15:03] <c_14> https://trac.ffmpeg.org/wiki/How%20to%20concatenate%20(join,%20merge)%20media%20files
[15:03] <daedeloth> awesome, thanks
[15:12] <nikita_> Hi! How to set periodic intra refresh with libav c/c++ api ?
[15:13] <nikita_> I mean x.264 encoder
[15:13] <Fjorgynn> what
[15:14] <nikita_> Fjorgynn, what - is it to me?
[15:17] <barry> does ffmpeg support encoding/decoding mlp
[15:21] <Fjorgynn> barry: ffmpeg -codecs
[15:21] <Fjorgynn> and there is mlp
[15:21] <barry> is that for both encoding and decoding
[15:21] <Fjorgynn>  D.A..S mlp                  MLP (Meridian Lossless Packing)
[15:22] <c_14> decoding, audio, lossy
[15:22] <c_14> s/y/less
[15:23] <barry> what about encoding?
[15:23] <c_14> Not with my configuration.
[15:23] <Fjorgynn> barry: you must compile that into it I think
[15:23] <barry> i just did  ffmpeg -i test.mlp test.wav    and  both file size is same. why is that
[15:23] <Fjorgynn> but not the default binary
[15:24] <c_14> > Code Status: MLP encoder never finished, MLP decoder committed to FFmpeg.
[15:25] <Fjorgynn> never heard about mlp btw
[15:25] <Fjorgynn> barry: mlp is lossless
[15:25] <barry> lossless audio codec
[15:25] <Fjorgynn> that's probably why it has the same filesize?
[15:25] <barry> no, test.wav should be way bigger
[15:26] <c_14> can you ffprobe both files and pastebin the output?
[15:26] <barry> Stream mapping:
[15:26] <barry>   Stream #0:0 -> #0:0 (mlp -> pcm_s16le)
[15:26] <barry> Press [q] to stop, [?] for help
[15:26] <barry> size=   21678kB time=00:00:57.80 bitrate=3072.0kbits/s
[15:27] <barry> why is it converting to  16bit only?
[15:28] <c_14> probably because you didn't specify a codec and that's the default that ffmpeg chose
[15:28] <Fjorgynn> yeah default preset?
[15:28] <barry> what is the right command then
[15:29] <barry> ffmpeg -i test.mlp test.wav
[15:29] <c_14> What are you trying to do?
[15:29] <barry> convert mlp to wav
[15:30] <Fjorgynn> and what bit?
[15:30] <Fjorgynn> 16,32,64?
[15:30] <barry> same and original
[15:30] <barry> same as original
[15:30] <Fjorgynn> try -sameq
[15:30] <barry> and it's 24bit
[15:31] <Fjorgynn> https://trac.ffmpeg.org/wiki/audio%20types
[15:32] <Fjorgynn> or somethink like ffmpeg -i test.mlp -c:a pcm_blueray test.wav
[15:32] <barry> i get error message
[15:33] <Fjorgynn> or somethink like ffmpeg -i test.mlp -c:a pcm_s24le test.wav
[15:33] <Fjorgynn> barry: my car doesn't start. It's red.
[15:34] <barry> what is the difference between LE and BE
[15:35] <Fjorgynn> good question
[15:35] <c_14> little endian and big endian
[15:35] <barry> and what does that mean
[15:35] <c_14> the order in which bits are arrayed
[15:36] <barry> what is the difference between signed and unsigned
[15:37] <c_14> signed numbers have negative values unsigned do not
[15:39] <barry> i don't get it
[15:39] <barry> so what is better
[15:40] <c_14> pick whichever sounds coolest, it doesn't really matter. They're all lossless.
[15:40] <Fjorgynn> why are you converting mlp to wave anyway?
[15:41] <barry> because then can convert to any other format
[15:43] <barry> okay ffmpeg -i test.mlp -c:a pcm_s24le test.wav  worked but what if i didn't know original file is pc_s24le
[15:43] <Fjorgynn> try -sameq
[15:44] <Fjorgynn> dunno
[15:44] <Fjorgynn> barry: you can always check what sort of audio you have
[15:44] <barry> sameq doesn't work
[15:44] <Fjorgynn> with ffmpeg or ffprobe
[15:45] <Fjorgynn>  Stream #0.0: Audio: mp3, 44100 Hz, stereo, s16, 256 kb/s
[15:45] <Fjorgynn> is what I get
[15:46] <barry> Input #0, mlp, from 'test.mlp':
[15:46] <barry>   Duration: N/A, start: 0.000000, bitrate: N/A
[15:46] <barry>     Stream #0:0: Audio: mlp, 96000 Hz, stereo, s32
[15:47] <barry> it's wrong, it's not 32
[15:49] <barry> is that a bug
[16:28] <daedeloth> has anyone played with ffmpeg and nodejs? (and maybe even node-webkit)
[16:41] <bornpilot> problem with fade out start time. I have video that is 30.83 min or 1800.83 secs I am trying to apply a fade out one second before the video ends for a duration of one second so the start time should be 1799.83 when I apply the following command it actually fades out at 1360 seconds what am I doing wrong?? ffmpeg -y -i $first -s 480x272 -bufsize 600 -minrate 0 -maxrate 600 -preset ultrafast -af "afade=t=out:st=$sec:d=1" -vf "fade=t=out:st=
[16:55] <DeadSix27> what is the most safe way to append 1 flac file to another without re-encoding?
[17:11] <bornpilot> try cat
[17:11] <DeadSix27> ye i used concat
[17:11] <DeadSix27> guess it was able to handle 5.1 flac and 2.0 flac together
[17:12] <DeadSix27> on the end it kept the 6channels, and appended a 2 channel file
[17:12] <DeadSix27> which has muted 4 channels i assume.
[17:44] <ReactorScram> I'm trying to use ffmpeg and x264 to record video from inside my OpenGL game
[17:44] <ReactorScram> However when I play back the video in mplayer, it pauses and stutters about every 700 ms
[17:45] <bornpilot> lower your bitrate
[17:48] <ReactorScram> I think my video is about 10 seconds long, and the file is 2,299,176 bytes. So it's 1800 kbit/s, and that's too high to play?
[17:51] <ReactorScram> If I write each frame to disk as a BMP, then encode with the ffmpeg CLI, it plays fine, so I believe the frames themselves are captured right
[18:00] <DeadSix27> bornpilot: also nvm i ran into issues
[18:00] <DeadSix27> bornpilot: it ignores time
[18:05] <ReactorScram> When I play it in VLC it doesn't pause but it appears to be at a very low FPS, and the debug log shows a lot of "picture is too late to be displayed" / "picture might be displayed late". I must have messed up the PTS / DTS somehow?
[18:15] <ReactorScram> Does the data in AVPacket need to be padded?
[18:21] <bornpilot> DeadSix27 do you need all the six channels on flac 5.1?
[18:21] <DeadSix27> yes
[18:22] <DeadSix27> bornpilot, i used to encode raw into flac and then append em
[18:22] <DeadSix27> would appending them while encoding them to flac
[18:22] <DeadSix27> fix the issue?
[18:24] <bornpilot> if the sources have different audio channels you could try to map them
[18:32] <DeadSix27> bornpilot what said solved my issue
[18:32] <DeadSix27> time codes are correct now
[18:32] <DeadSix27> apparently ffmpeg handled it itself
[18:32] <DeadSix27> i dont rly know how, nor does it matter.
[18:32] <DeadSix27> -> [dca @ 028baec0] Number of channels changed in DCA decoder (6 -> 2)
[18:32] <DeadSix27> but muxed right and plays right
[18:34] <bornpilot> cool
[18:37] <DeadSix27> also i mean not mapping
[18:38] <DeadSix27> i mean just using concat on the raw rather than after (didnt do that before due to lazyness)
[20:03] <DeadSix27> what does the (native) next to PCM mean btw?
[20:03] <DeadSix27> that ffmpeg supports it without any codec? e.g it says (flac) behind flac, meaning it uses the flac de/encoder ?
[20:15] <benlieb> how do I know what the default codec used when I just do something like ffpmpeg -i input output ?
[20:16] <c_14> ffmpeg -h muxer=$muxer
[20:16] <benlieb> hi c_14 :)
[20:17] <c_14> hi
[20:17] <benlieb> that for me?
[20:17] <c_14> yep
[20:18] <benlieb> ah that works
[20:18] <benlieb> turns out that change in file size was because the orig had two audio tracks..., and one got removed, and the other reduced.
[20:19] <benlieb> or rather went from 154 kb/s to 98 kb/s
[20:19] <benlieb> c_14: ^
[20:23] <benlieb> c_14: are there docs to understand the output of ffprobe?
[20:23] <benlieb> for example this line seems pretty important, but I only understand a little bit of it: Stream #0:0(und): Video: h264 (High) (avc1 / 0x31637661), yuv420p, 640x362 [SAR 1:1 DAR 320:181], 269 kb/s, 29.97 fps, 29.97 tbr, 30k tbn, 59.94 tbc (default)
[20:23] <c_14> There's the manpage and the wiki page.
[20:23] <benlieb> well half
[20:24] <benlieb> c_14: this page just discusses options not output: http://www.ffmpeg.org/ffprobe.html
[20:27] <c_14> hmm, you're right there's no good output doku. Or I can't find it at least.
[20:27] <c_14> What parts of that line don't you understand?
[20:27] <benlieb>  (avc1 / 0x31637661), yuv420p
[20:27] <benlieb> (und)
[20:28] <benlieb> 29.97 tbr, 30k tbn, 59.94 tbc (default)
[20:29] <c_14> The (und) is usually where the language code goes.
[20:29] <benlieb> This command (ffmpeg -i 1.mp4 -vf scale=iw*sar:ih -c:a copy -map_chapters -1 1_scaled.mp4) is changing my video stream from
[20:29] <benlieb> Stream #0:0(und): Video: h264 (High) (avc1 / 0x31637661), yuv420p(tv, smpte170m), 720x362 [SAR 8:9 DAR 320:181], 311 kb/s, 29.97 fps, 29.97 tbr, 90k tbn, 180k tbc (default)
[20:30] <benlieb> to
[20:30] <benlieb> Stream #0:0(und): Video: h264 (High) (avc1 / 0x31637661), yuv420p, 640x362 [SAR 1:1 DAR 320:181], 269 kb/s, 29.97 fps, 29.97 tbr, 30k tbn, 59.94 tbc (default)
[20:30] <benlieb> so the the kb/s changes as well as the tbn, and tbc
[20:30] <benlieb> whatever that is
[20:31] <benlieb> and also the yuv420p
[20:31] <c_14> the kb/s is because of the video being reencoded, you can mess around with that in the options
[20:31] <benlieb> yeah figured that one
[20:32] <c_14> yuv420p is the pixel format the (tv, smpte170m) are the colormatrix
[20:34] <c_14> tbn is the time base in AVStream from the container, tbc is the time base in AVCodecContext for the codec for the stream, tbr is guessed from the video stream and is the value users want to see when they look at the video frame rate
[20:35] <benlieb> a time base of 30k means 30k of what?
[20:45] <c_14> The timebase is just the reference number in terms of which the timing information is represented. And it's actually the inverse of what ffmpeg says.
[20:45] <c_14> ie timestamp * time_base = 'real' seconds
[20:46] <benlieb> c_14: sounds like something I don't need to think about as it relates to file size or quality, or should I be concerned about this?
[20:46] <c_14> Not unless you're building your own decoder.
[21:03] <benlieb> c_14: as you said this (ffmpeg -i 1.mp4 -vf scale=iw*sar:ih -c:a copy -map_chapters -1 1_scaled.mp4) sets my SAR to 1:1
[21:05] <benlieb> from 720x362 [SAR 8:9 DAR 320:181]  to =>  640x362 [SAR 1:1 DAR 320:181]
[21:05] <benlieb> but when I add the crop (ffmpeg -i 1.mp4 -vf scale=iw*sar:ih -c:a copy -map_chapters -1 -vf crop=w=640:h=360 1_scaled.mp4), it goes right back
[21:06] <benlieb> 640x360 [SAR 8:9 DAR 128:81]
[21:06] <benlieb> the video doesn't display warped though, it displays about 10% less wide
[21:08] <c_14> you can't have multiple -vf commands
[21:08] <c_14> try -vf scale=iw*sar:ih,crop=w=640:h=360
[21:08] <benlieb> ooh I didn't mean to
[21:10] <benlieb> lets see if that fixes it
[21:10] <benlieb> otherwise I could try setdar=dar=16/9
[21:15] <benlieb> yay that fixed it
[21:15] <benlieb> happy day
[21:23] <benlieb> c_14: could you explain why the "video" has 436 kb/s nitrate, but the video stream says 272 kb/s. What is the latter a measure of?
[21:23] <benlieb> bitrate
[21:24] <c_14> video bitrate + audio bitrate?
[21:25] <benlieb> yeah that looks right
[21:26] <benlieb> c_14: so my video nitrate is changing during scale, but I don't see any options to maintain it, or specify it...
[21:27] <benlieb> bitrate damn spellcheck
[21:27] <c_14> Are you using libx264?
[21:28] <benlieb> I'm using the default, and that looks like that
[21:28] <benlieb> Mime type: application/mp4.    Default video codec: h264.
[21:29] <c_14> https://trac.ffmpeg.org/wiki/Encode/H.264
[21:29] <c_14> Basically pick a preset and a crf you're happy with.
[21:36] <benlieb> that's the most useful page ever
[21:36] <benlieb> looks like the default cry is 23 and preset is medium
[21:36] <benlieb> crf
[21:36] <benlieb> gr spellcheck
[21:37] <benlieb> what qualities would you recommend for a site that offers instructional videos?
[21:37] <benlieb> video qualities, resolution, crf, etc?
[21:37] <benlieb> I'm going to have to send you a beer or a gift in the mail :)
[21:38] <c_14> Are the video sources raw or have they been encoded before?
[21:38] <sacarasc> What kind of instructional videos? How to do watch repair? Or something more like how to do basic bricklaying? Or how to make a GIF in Photoshop?
[21:38] <benlieb> c_14: they will be coming mostly from DVD's
[21:38] <benlieb> They are dance instructional videos
[21:39] <sacarasc> You probably want a farily low CRF, then, at least for the faster dances.
[21:39] <sacarasc> *fairly
[21:39] <c_14> I'd probably stay as close to the source as possible then.
[21:39] <benlieb> I was running a site as web dev with 3 other people. But they decided to move on after 5 years. I never handled any of the video.
[21:40] <benlieb> All the video was done manually by another guy, which "worked" but the video wasn't shot well, and the process of adding content wasn't very automated. If I want to keep the business I have to figure out video as well as web dev, and see how much I can automate.
[21:41] <benlieb> First step for me is to understand video and DVD conversions. I probably won't be doing any filming any time soon, but that would be cool at some point.
[21:41] <c_14> ie try to keep the resolution as close to the source as you can, and start with a crf around 20 and then modify as necessary.
[21:44] <benlieb> the default video codec is h264. Do I have to specify that if I want to use the presets and crf ?
[21:44] <c_14> nah
[21:44] <c_14> I usually do though, just for clarity when I look at my commandline history.
[21:45] <benlieb> yeah, good idea
[21:45] <benlieb> c_14: what are you using video for?
[21:48] <c_14> Ripping dvds and blurays I buy into digital form. Some home video stuff. Fun.
[21:49] <benlieb> c_14: you sure know a lot about video for a hobbyist
[21:50] <benlieb> do I put -c:v libx264 -preset slow -crf 22 before or after filter?
[21:51] Action: c_14 prefers after but isn't sure it makes a difference
[21:57] <whitepearl> Hi, Using 32 bit latest release of ffmpeg(http://ffmpeg.gusari.org/static/) I am unable to use stereo3d filters .. Error: "[AVFilterGraph @ 0x9c90080] No such filter: 'stereod3d' Error opening filters!"
[21:57] <whitepearl> Can anyone help?
[21:58] <whitepearl> Also tried using earlier versions, do I need to enable filters?
[21:58] <benlieb> whitepearl: when that kind of thing happens to me, I usually have to recompile ffmpeg with the flag that grabs the thing I don't have
[21:58] <sacarasc> whitepearl: Is it because you typo'd it?
[21:58] <benlieb> I've had to do it 5 times in the a last two weeks
[21:58] <sacarasc> Looks like you have an extra 'd' in there.
[21:59] <whitepearl> I did not  have to compile it from source.. it directly gave me binaries
[22:00] <whitepearl> http://pastebin.com/skEPV5ty Config
[22:00] <benlieb> whitepearl: did you typo?
[22:00] <benlieb> what system are you on?
[22:00] <whitepearl> rhel 6
[22:00] <benlieb> I'm only familiar with mac via brew
[22:01] <whitepearl> tried on mac also
[22:01] <whitepearl> doesn't works
[22:01] <benlieb> did you spell it wrong like sacarasc says?
[22:01] <benlieb>  stereod3d != stereo3d
[22:05] <whitepearl> oh ya.. it works! Thnx guys
[22:08] <benlieb> whitepearl: nice
[22:08] <benlieb> now do 10 pushups :)
[22:09] <benlieb> c_14: what do you think a reasonable resolution (size) would be for an online video product?
[22:09] <benlieb> we want to maximize quality and minimize storage fees and load times
[22:10] <theekoz> anyone know how to fix/set PCR to more than 40ms
[22:10] <theekoz> or how to flag every I Frame
[22:11] <c_14> benlieb: Probably either 480 or 720p.
[22:11] <benlieb> c_14: somehow I ended up with 640x360, which I guess is in the middle
[22:12] <benlieb> I think that was the res on the last DVD I imported, which was somewhat of an amateur product, and not high res
[22:15] <sacarasc> benlieb: 640x360 would be 360p (if progressive, 360i if interlaced).
[22:15] <benlieb> what does the p mean?
[22:15] <sacarasc> Progressive.
[22:15] <benlieb> i'm such a noob
[22:16] <sacarasc> No interlacing, basically.
[22:16] <benlieb> 640x360 is a resolution, what is 360p a measure of?
[22:17] <sacarasc> It says it's resolution (approximately) and that it is progressive.
[22:18] <beastd> benlieb: partial resolution with information about the video content being progressive or interlaced
[22:18] <benlieb> so just the height?
[22:19] <benlieb> or is is some formula?
[22:19] <c_14> just the height, the width depends on the aspect ratio
[22:20] <benlieb> ok so my vids are a bit on the lower "quality" spectrum for web videos for this particular set of dvds
[22:44] <benlieb> what is an appropriate resolution for mobile videos. We offer both versions for the desktop, and for mobile
[22:50] <benlieb> c_14: ^
[22:52] <benlieb> sacarasc: ^
[22:53] <c_14> I'm not very experienced in the mobile sector. Modern mobile devices have similar resolutions to monitors though so maybe just the same?
[22:54] <benlieb> c_14: I guess I mean what is the smallest resolution size I could use that would not look bad
[22:54] <benlieb> I guess I could do some tests :)
[22:56] <benlieb> is the -s option the same as scale filter?
[22:56] <c_14> ye
[22:56] <c_14> -s is the equivalent as appending the scale filter to a filterchain
[23:32] <benlieb> can something encoded at lower resolution "make up" quality in bitrate?
[23:33] <benlieb> basically are bitrate and resolution on a fluid spectrum like image resolution and ppi? So that something saved at a high ppi with lower resolution still can appear as higher quality on screen?
[23:36] <c_14> ye
[23:40] <iive> yes. if you have target bitrate and you can't improve the quality by tuning the codec, then you lower the resolution.
[00:00] --- Sun Jun  8 2014


More information about the Ffmpeg-devel-irc mailing list