[Ffmpeg-devel-irc] ffmpeg.log.20140225

[burek]

[02:06] <Keshl> Is there a magical set of settings I should use for H.265 encoding? It "works", but the video's of.. Very, questionable, quality.. XD
[02:08] <Keshl> I tried "ffmpeg -i rawuncompressedfootagehere.flv -vcodec libx265 -preset medium -crf 18 -movflags faststart -pix_fmt yuv420p -vf scale=1366:768 -acodec copy blah.mkv     I don't care which container is used. MKV was just the first to come to mind after MP4.
[02:36] <iive> Keshl: where did you got libx265 ? (5 as in Five) ;)
[02:36] <Keshl> Newest builds have support built in, oÉo.
[02:37] <Keshl> They have since the 11th, I think.
[02:37] <Keshl> In my case, I got'd from http://ffmpeg.zeranoe.com/builds/ -É-.
[02:45] <Keshl> WHOA °²°
[02:45] <Keshl> Figured it out.
[02:45] <Keshl> And and whooooooa o_o
[02:46] <Keshl> Uh, right
[02:46] <Keshl> ffmpeg -i awesomevideohere.flv -x265-params crf=18 -movflags faststart -pix_fmt yuv420p -vf scale=1366:768 -acodec copy blah.mp4
[02:47] <Keshl> Yes, you do *not* specify -vcodec. -x265-params does that, appearntly.
[02:49] <Keshl> I personally see a loss in quality. It's extremely minor, though, and the filesize is redced by 50% even with highly unpredictable motion scenes.
[02:49] <Keshl> For a 50% reduction with 264 the loss of quality's far more noticable.
[03:01] <Keshl> Uh, the fudge?
[03:02] <Keshl> Maybe more changed during the update than I thought. I was on a fairly old build.
[03:02] <Keshl> I noticed changing the CRF in 265 didn't change the videos at all. Same appearnt quality, same output filesize.
[03:02] <Keshl> For kicks I ran a CRF of 23 in 264. Same appearnt quality as 265, and same exact filesize.
[03:02] <Keshl> Default CRF if not supplied is 23.
[03:02] <Keshl> So, what's going on here? o_o
[03:05] <relaxed> Keshl: try -preset veryslow
[03:06] <relaxed> also pastebin.com
[03:06] <Keshl> ... D'oh.
[03:07] <Keshl> Hold on a moment. Think I see what I did wrong. <É<'
[03:07] <Keshl> Nope. Seems that as soon as I specify -x265-params it just ignores any -preset or -crf I give it.
[03:08] Action: Keshl pastes.
[03:08] <relaxed> well, support for libx265 is fairly new.
[03:09] <Keshl> relaxed: http://pastebin.com/zuhRH4uG
[03:09] <Keshl> I know. I think I found a bug. D:
[03:09] <Keshl> Either that or I'm horribly misunderstanding the syntax of the command.
[03:09] <Keshl> Note: Windows is stupid and linewraps at 80 characters. I didn't actually put that command on 3 lines.
[03:10] <Keshl> .... Oh.
[03:10] <Keshl> "Codec AVOption x265-params (set the x265 configuration using a :-separated list of key=value parameters) specified for output file #0 (E:\crf_crf_265_18.mp4) has not been used for any stream."
[03:10] <Keshl> So it /is/ ignoring what I tell it.
[03:11] <Keshl> That still doesn't explain why the filesize for 264 and 265 are completely identical down to the kb.
[03:11] <Keshl> And why the apparent visual quality of the frames are identical, assuming a crf of 23.
[03:12] <Keshl> Really seems like it's using the same encoder, and that encoder seems to be 264.
[03:13] <relaxed> Stream #0:0 -> #0:0 (h264 -> libx264)
[03:13] <relaxed> add -c:v libx265
[03:13] <Keshl> ... How'd I miss that?!
[03:13] <Keshl> Should I add it before or after -x265-params?
[03:14] <relaxed> before gives me the warm fuzzies
[03:14] <Keshl> oÉo.
[03:14] Action: Keshl snugs dem warm fuzzies. -É-.
[03:15] <Keshl> Yes, now it's using 265 for sure cuz it won't write to .mp4..
[03:15] Action: Keshl tries MKV.
[03:16] <Keshl> MKV works. Encoding is terrifyingly slow. This is encouraging. <É<
[03:22] <Keshl> ... Uh.
[03:22] <Keshl> The output file's 28 bytes.
[03:22] <Keshl> Wait.
[03:22] <Keshl> brain.
[03:22] <Keshl> Got it.
[03:23] <Hello71> -map 0:
[03:23] <Keshl> Colors are a little off, but the fidelity's what I expect from crf 18 now.
[03:23] <Keshl> Sorry, silly my played the bad mp4 first. <.< MKV's.. "Fine".
[03:23] <Keshl> Size's far larger than 264, though. 50% larger.
[03:24] <Keshl> Trying CRF23, just to see if the quality's the same and CRF needs rebalancing..
[03:36] <Keshl> Okay, 17.9 megs. 15-20% smaller. Video quality is better than 264 crf23 but worse than 264 crf18, and in different areas. I get that that's by design, but I felt it's worth saying.
[04:29] <Keshl> In h.265, CRF22 is roughly equal to h.264's CRF18. The quality's visually identical from what I can tell apart from some momentary stutters at the start of the video, but the difference in filesize is amazing.
[04:30] <Keshl> First, let me say that, at 23 CRF, the default for 264 and 265, the video I have comes to 22.1 megs.
[04:30] <Keshl> At a CRF of 22 in 265 (Which looks like a CRF of 18 in 264), the file is 20.4 megs.
[04:30] <Keshl> A CRF of 18 in 264 is 38.3.
[04:31] <Keshl> So, essentially, 265's become "free" visually lossless video, as far as I'm concerned.
[04:32] <Keshl> And it turns out the bad colors are due to me playing 265 files in another player. 264 files are discolored the same way there, so it's the decoder at fault.
[04:36] <Keshl> Playing with the usual player (Updated) does fix the color issues, and fixes the skipping. By replacing it with a black screen.
[04:37] <Keshl> Regardless. Shiny. -É-.
[05:46] <Keshl> Wow, encoding's only about four to five times as long as 264.. XD
[06:07] <Keestu> hi all, i have a basic question. how can i find how many frames per second i am receiving by programatically ?
[06:08] <Keestu> for example,  if fps is said to be 25, then does this av_read_frame() retrieves 15 frames per second ?
[06:21] Action: Keshl snugs a Zeranoe cuz builds. -É-.
[06:26] <Dark-knight> If i use "-c copy" in the command line, doesn't to do the same thing as putting "-acodec copy -vcodec copy" in the command line?
[06:26] <Dark-knight> also if it does, what is the difference?
[06:44] <Dark-knight> and when changing the container (for ex: from avi to mp4) could i just leave out the copy commands completely? had just have a code that looks like this?
[06:44] <Dark-knight> ffmpeg -i input.avi output.mp4
[06:51] <LithosLaptop> dont think so
[06:53] <Dark-knight> don't think what?
[06:56] <Zeranoe> Keshl: Whats up?
[07:05] <Keshl> Zeranoe: Jus'huggin. I 'preciate the builds and never knew'ya came here. oÉo.
[07:05] <Keshl> Dark-knight: No, ffmpeg is stream-based. You'd have to reencode while changing containers. Pretty sure, anyway.
[07:06] <Dark-knight> so i would have to use a copy command?
[07:07] <Keshl> At least.
[07:07] <Dark-knight> now back to my question. which copy command should i use?
[07:07] <Keshl> -vcodec copy -acodec copy
[07:07] <Dark-knight> I have used -c copy and it did the same thing. what is the difference?
[07:07] <Keshl> Do note that not all containers support all formats.
[07:07] <Keshl> Probably nothing. Just shorthand.
[07:08] <Keshl> There's -c:v that you can use instead of -vcodec too.
[07:08] <Dark-knight> so is -c copy the same code just shortened?
[07:09] <Keshl> Possibly. I don't know for sure, but if it seems to do the same thing, then it could be.
[07:09] <Dark-knight> i assume that -c copy is a combination of -vcodec copy -acodec copy
[07:14] <Dark-knight> I need to confirm if that is true
[07:16] <Keshl> Well, wait until someone else comes around.
[07:17] <Keshl> Remember that we are not paid to help you, by you or by anyone else. We all have lives. Be patient. You are not automatically entitled to free support at any point. If you want help, you gotta act nice.
[07:21] <bparker> or you could just try it yourself
[07:22] <Keshl> Indeed. <É<
[07:23] <Dark-knight> i understand that, if you noticed, I didn't demand anybody to immediately answer my question. I simply made anyone reading this aware that I needed confirmation on my clam. So before you spout things at me about being unpatient understand I didn't provide any sense of urgency in my statement.
[07:23] <Dark-knight> please refrain from such statement in the future
[07:25] <Keshl> You /did/ provide a sense of urgency. You didn't intend to, but it's what you did.
[07:25] <Keshl> On top of that, you answered your own question and told us that, twice, and then asked us for help on it anyway.
[07:25] <Keshl> What do you expect us to do? o.O'
[07:31] <Keestu> can someone kindly help me, regarding this frame per second concept?.
[07:31] <Keshl> ... I wonder if -preset placebo works with 265. That'd be painful. x.x With Medium on an i7 I'm getting 2.6 FPS.
[07:31] <Keshl> Keestu: What was it, oÉo?
[07:32] <zap0> Keestu, there are these things called 'frames'.  somewhere in the header, a value is set called 'fps'.   players use this as a guide to the rate they display them.
[07:34] <Keestu> thanks for the reply  keshl, i am  a newbie and trying to understand the concepts behind ffmpeg/video.   fps -> frame per second , for example 25fps -> 25 frames are captured in a second. Now i  have written a simple video player in ffmpeg to display.  i read frames by av_read_frame(),  who can i related this 25 fsp to this API?
[07:35] <Keestu> that meas ,av_read_frame() , does it  require to read 25 fps ?
[07:35] <Keshl> That's way beyond my skillset, sorry D: I just "kinda sorta" know how stuff works, and just barely enough to know what some of the more important paramaters do when I'm encoding stuff.
[07:35] <Keshl> And until this second, I didn't even know ffmpeg had an API.
[07:37] <Keestu> Keshl,  no issues, you helping intention really appreciable . :)
[07:37] <Keestu> if anyone knows the link also, kindly share with me, i can read the same and try to understand.
[08:22] <Dark-knight> I understand that putting "-vcodec copy -acodec copy" in the command line will produce the same video as putting "-c copy" in the line, BUT causation doesn't always mean correlation. so I was wondering if "-c copy" is in-fact a shortened code for "-vcodec copy -acodec copy" or if the videos are slightly different because of the different code?
[08:23] <Dark-knight> i can't tell
[08:23] <relaxed> -map 0 -c copy means copy everything
[08:24] <relaxed> get in the habit of mapping what you want
[08:25] <relaxed> for just the audio and video streams, ffmpeg -i input -map 0:v -map 0:a -c copy ...
[09:22] <Keestu> hi all, avformat_open_input () -> when i tried to open with rtsp://<ip>:<port> , it works in all mobiles [htc, samsung], but when i change to http://<ip>/test.sdp , it works only with samsung, but not in HTC.  In which context can i look this issue ?
[10:10] <Waster> hey all, is it possible to have multiple encodings with ffmpeg with one of these recording to the file also?
[10:16] <Waster> ffmpeg -i <RTSP_URL> -c:v libx264 -b:v 128k -f flv rtmp://localhost:1935/hls/stream_low -c:v libx264 -b:v 256k  -f flv rtmp://localhost:1935/hls/stream_mid -c:v libx264 -b:v 512k -f flv rtmp://localhost:1935/hls/stream_hi
[10:16] <Waster> it works well
[10:16] <Waster> but I want to record last 512k encoding to the file also
[10:17] <Waster> -f tee for the last stream does not work
[10:35] <Waster> anyone?
[11:05] <Keshl> I now understand why hardware decoding is a wonderful thing.
[11:05] <Keshl> Turns out, no, a single core at 3.6 ghz cannot handle playing 1920x1080 video at 60 FPS when encoded with h.265.
[11:05] <Keshl> That was a very bad idea.
[11:05] <Keshl> XD
[11:09] <BtbN> well, that'll improve a lot in the future
[11:09] <BtbN> the code isn't yet optimized
[11:19] <Keestu> Keshl,  may  i ask you how to do harware decoding?, i am aware of software encoding with the help of libx264.
[11:19] <Keshl> Keestu: Hardware decoding happens when playing video back in a media player. To my knowledge, ffmpeg doesn't do it.
[11:20] <Keestu> ok, that means media player uses hardware encoding ?
[11:24] <Keshl> Depends on the player. Most do.
[11:29] <Mavrik> there are also several layers of hardware acceleration
[11:34] <Keshl> ... I'm just full of bad ideas today. XD
[11:34] <Keshl> So I tried using ultrafast rather than medium since ultrfast's usually easier for my system to decode.
[11:34] <Keshl> Bitrate: 144000ish.
[11:35] <Keshl> ... I don't think this is a good idea anymore. XD
[11:38] <Keshl> Man. This reminds me of the days when I was still running an nVidia card with four digits and trying to play h.264 at 1080/60. It just wasn't happening. XD Doesn't help that no GPU supports h.265..
[11:42] <esperegu> Hi. I have a annoying noice on an audiostream. anyone knows how I can get rid of it? Is there a way to filter out noice or maybe play nothing unless above a certain amount? thx!
[11:49] <Keshl> esperegu: it depends on the sound, speciffically. Some are easier to take out than others.
[11:49] <esperegu> Keshl: you have any commands for me I can test?
[11:50] <Keshl> esperegu: Start by extracting the audio from the video. Run this, substituting filenames where required:
[11:51] <Keshl> ffmpeg -i myvideo.avi audio.wav   (Make sure you use .wav)
[11:51] <Keshl> Then play the audio, make sure it actually plays. If it does, good.
[11:51] <Keshl> next, grab http://audacity.sourceforge.net/ and open it there.
[11:52] <Keshl> Depending on the specific type of sound you're hearing, the next step may be dead simple or it may require some experimentation.
[11:53] <Keshl> If it's something along the lines of a high-pitched squeal, just import the .wav file (FIle > Import > Find it > Just hit OK), then do Effect > Low Pass filter. Set the rolloff to 48db and the cutoff to 16000, hit okay, be patient, then play it and see if you still hear it.
[11:54] <Keshl> If you still do but it's quieter, undo (Ctrl Z) and try again but with a lower cutoff. If you bring it too low you'll eventually notice it affect more of the audio than you want it to, so that might take some experimentation.
[11:54] <Keshl> If it has no effect, then you gotta do something a bit more complicated. It "should" work, provided you spend some time with it, though.
[11:55] <esperegu> and then howto get it into ffmpeg?
[11:55] <Keshl> First, find a part of the audio that has /just/ the sound you want to remove and nothing else.
[11:55] <Keshl> (I'll explain that in a minute.)
[11:55] <esperegu> Keshl: thats no problem cause its a noice playing all the time
[11:56] <Keshl> Once you've found it, select it (Just click and drag over it), then do Effect > Noise Removal, and click "Get Noise Profile". The window'll just dissappear and look like nothing happened.
[11:56] <Keshl> next, select everything (Ctrl A), then run Noise Removal again. Instead of hitting that first button though, just hit OK. This might take a while for it to process depending on the complexity of the noise, but eventually it'll finish.
[11:57] <Keshl> Once it does, play and see if you still hear it. If you do but it at least sounds a little different, select another part of audio that has just the noise and repeat both steps until it's all gone.
[11:57] <Keshl> It's rare you'll get rid of it /all/ on your first try, but there is a chance you'll get lucky.
[11:57] <Keshl> Regardless, whenever it's cleaned up enough for'ya, export it as .wav again (You can overwrite the same file if you want, just make sure you /export/, not /save/).
[11:58] <Keshl> Then in ffmpeg, do this:   ffmpeg -i videowithnoise.avi -i fixedaudio.wav newvideo.avi       Yes, you specify -i twice. I don't know if that's /exactly/ what you do to fix it, but I know it's something along those lines.
[12:00] <Keshl> I can't check right now since I'm busy encoding a video myself and if I try to start another up my computer's gunna scream bloody murder at me. <.< But if that doesn't work, try:   ffmpeg -i video.avi -i fixed.wav -vcodec copy newvideo.avi   or something.
[12:01] <Keshl> (It's possible you'll need to use "-map". I have _no_ clue how this switch works or what you'd need to type to make it replace audio. Hopefully somebody else does..)
[12:05] <esperegu> Keshl: k. trying it. if you interested its the noice in the beginning of this file: http://babilejo.org/misc/noise2.wav
[12:06] <Keshl> Oh, that hiss.
[12:07] <Keshl> Yeah, got rid of it. That one's rather easy. XD
[12:07] <Keshl> If it's something like a crowded subway it's harder but still doable. This though, first try, second method.
[12:07] <esperegu> that be awesome
[12:08] <Keshl> Select some of the noise, Effect > Noise removal, get noise profile, select all, effect > Noise removal, hit OK, done.
[12:08] <Keshl> (By "some" I mean "1-2 seconds")
[12:08] <esperegu> Keshl: in you example you join them again.
[12:08] <Keshl> It works?
[12:09] <esperegu> Keshl: this can't be done by ffmpeg?
[12:09] <Keshl> What can't be done, exactly? Join what things?
[12:09] <esperegu> Keshl:  ffmpeg -i videowithnoise.avi -i fixedaudio.wav newvideo.avi
[12:10] <esperegu> Keshl: remove the noice
[12:10] <esperegu> Keshl: I am looking for an option so ffmpeg can remove that hiss
[12:11] <Keshl> esperegu: It can't.
[12:11] <Keshl> It's for video processing, not audio processing.
[12:11] <Keshl> The most ffmpeg can do with audio, as far as I know, is transcode it (from like, wav to mp3 and such) -- No actual effects.
[12:11] <JEEB> there are audio filters
[12:12] <Keshl> Oh, shiny. -É-.
[12:12] <Keshl> JEEB: Any chance you know if the command I'm giving him is right, and it'll overwrite the audio stream already in the .avi?
[12:13] <JEEB> it most definitely isn't correct without -c:v copy IMHO (since audio has to be re-encoded again unless you want uncompressed audio in that avi) :P
[12:13] <esperegu> Keshl: its not an avi but a stream but I suppose its the same
[12:14] <JEEB> also mapping is good
[12:14] <JEEB> ffmpeg cli app will select of course one audio and video track by default, but you really can't know if you're getting the right one otherwise :P
[12:14] <esperegu> isn't there a filter that plays no audio unless above a threshold or so?
[12:14] <Keshl> esperegu: I think there is, but then you'll hear the hiss when there's that instrument playing.
[12:15] <Keshl> If you did what I said, it'd become crystal clear. (It did for me.)
[12:15] <JEEB> feel free to go through the listing :P http://ffmpeg.org/ffmpeg-all.html#toc-Audio-Filters
[12:16] Action: Keshl tilts his head at "earwax"..
[12:20] <esperegu> Keshl: it is clear indeed. but if ffmpeg is not able to do it it does not really help
[12:20] <Keshl> Why not? o.O
[12:21] <esperegu> Keshl: it is a live stream
[12:21] <Keshl> Ohhh.
[12:21] <Keshl> Is there any reason you can't record the stream and process it later?
[12:22] <esperegu> Keshl: cause I listen to it real time ;-)
[12:22] <JEEB> anyways, go through the ffmpeg cli audio filters
[12:22] <JEEB> and see if you can use any of them
[12:22] <Keshl> JEEB: I looked. There's none. D:
[12:22] <esperegu> JEEB: I already tried a few but heard no difference. I might be using it wrong though
[12:22] <Keshl> esperegu: Why not just do it on your sound card, then?
[12:23] <Keshl> Just go to the equalizer and lower 16000 hz all the way or something.
[12:23] <JEEB> lowpass/highpass sound like something useful for denoising if that's what you want :P
[12:24] <esperegu> JEEB: I tried: -filter:v highpass
[12:24] <esperegu> but did not notice a difference
[12:24] <JEEB> > :v
[12:24] <JEEB> u w0t m8t
[12:24] <esperegu> JEEB: is the command correct?
[12:25] <JEEB> you are applying a filter to :v
[12:25] <JEEB> which is VIDEO
[12:25] <Keshl> JEEB: It's not a high-pitched or low-pitched sound.
[12:25] <JEEB> anyways, go read the fine documentation since it's there
[12:25] <JEEB> Keshl, that's not what high/lowpass does
[12:25] <Keshl> It's about 8,000 hz and it's just random mic static.
[12:25] <JEEB> unless I misunderstand
[12:25] <Keshl> You do. XD
[12:25] <Keshl> See, I've got sensitive ears. I actually /have/ to run my audio through a lowpass filter or I can't listen to 90% of stuff online.
[12:26] <Keshl> So believe me I know how it works.
[12:26] <Waster> is it possible to use tee and flv rtmp output?
[12:26] <Keshl> It basically does frequency analysis on the audio, and then removes audio that's higher or lower frequency than that from the audio stream.
[12:30] <Keshl> It /could/ work with a lowpass filter, but then it's sound like he's underwater. The noise ragnes from 5k to 8k, so if he cut down to 5k that'd be.. Intense. XD
[12:30] <Keshl> Usually if you cut down to even 12k you'll hear a bit of an underwater effect.
[12:31] <Keshl> And if he tried a highpass instead, he'd have to cut up to 8k. His audio only goes to 7k or so, so then it'd just be silence.
[12:32] <esperegu> isnt there a filter that cuts off everything under a certain level?
[12:32] <Keshl> If you want to see a visual demonstration, open up Audacity and any audio file. Click the little down arrow by the "mute" and "solo" buttons (Upper right corner of that section) and change it to a spectrogram.
[12:32] <Keshl> esperegu: There is not.
[12:32] <Keshl> esperegu: If you want to do something like that just for you, you can do that on your sound card instead.
[12:32] <Keshl> (provided you have one. A good one that actually supports stuff like that.)
[12:41] <esperegu> Keshl: its a bit more involved. ffmpeg streams my pulseaudio via rstp
[12:42] <Keshl> Oh, Linux? I have no idea how to get sound cards to do magic stuff there. Sorry. x.x
[12:44] <Waster> ok, solved it with -map 0 -flags +global_header
[12:44] <Waster> now tee works with rtmp
[12:44] <Keshl> Shiny -É-
[12:44] <Keshl> And thanks for telling us, oÉo
[12:45] <Waster> welcome...)
[12:45] <Waster> btw, is it possible to do daily archive then with ffmpeg?
[12:45] <Keshl> Can'ya describe what a "daily archive" is, oÉo?
[12:46] <Waster> I mean one recording per day with date in filename
[12:46] <Waster> automatically
[12:49] <Keshl> Assuming you know how to do some scripting (With Bash or powershell), sure.
[12:50] <Waster> sure I know, you mean I can create template for ffmpeg command line and run it through cron every day klling ffmpeg?
[12:50] <Keshl> Yes, that should be rather simple provided you know how to use Bash.
[12:50] <Keshl> I don't, though.
[12:51] <esperegu> Keshl: so there are no options for me?
[12:52] <Keshl> esperegu: There might be, but I don't know of any.
[12:52] <Waster> great, thanks
[12:53] <Keshl> Welcomes, oÉo
[12:54] <esperegu> JEEB: you have any idea?
[13:14] <Keshl> ... Blah D: The Linux version ain't compiled with libx265 xwx..
[13:43] <_aeris_> hello !
[13:43] <_aeris_> is there any option to force png output with image2 output format ?
[13:43] <_aeris_> if you specify an output file with png extension, ffmpeg generate a png
[13:44] <_aeris_> but with «-f image2 pipe:1», only jpeg :(
[13:48] <relaxed> _aeris_: try -c:v png
[13:51] <_aeris_> \o/
[13:51] <_aeris_> thanks relaxed
[13:54] <relaxed> Keshl: which linux version?
[13:55] <relaxed> _aeris_: you're welcome
[13:56] <Keshl> relaxed: The.. One?
[13:56] <Keshl> Any from http://ffmpeg.gusari.org/static/ .
[13:56] <relaxed> my build has it http://johnvansickle.com/ffmpeg/
[13:59] Action: Keshl pooookes..
[14:02] <Keshl> Nupe, dun gots. D:
[14:02] <Keshl> You have to compile with --enable-libx265 , oÉo.
[14:02] <JEEB> I will have to note though that x265 is very often broken in either compilation or in what it actually does
[14:03] <JEEB> also it is only useful if you do --preset placebo --ref 16 --bframes 16 level encoding
[14:03] <Keshl> Works fine for me on windows. D:
[14:03] <relaxed> Keshl: that build has support, trust me
[14:03] <JEEB> that's just because you're lucky
[14:03] <JEEB> (and --tune ssim since otherwise it optimizes for PSNR)
[14:03] <relaxed> I just tested it.
[14:03] <Keshl> relaxed: "Unknown encoder 'libx265'"
[14:03] <Keshl> I just tested it too. Fresh download.
[14:03] <Keshl> I'm using 26/five/, not 26/four/.
[14:03] <relaxed> are you using the 10bit binary?
[14:03] <relaxed> use the regular one
[14:03] <Keshl> No. No sane media player can play that. D:
[14:04] <JEEB> he wants to encode HEVC
[14:04] <Keshl> Using regular.
[14:04] <JEEB> not AVC
[14:04] <relaxed> my build has libx26FIVE support
[14:04] <Keshl> JEEB: Incredibly lucky, appearntly. The video I'm encoding is.. Scary-evil impossible-to-compress. XD
[14:04] <Keshl> relaxed: It really, truly, does not.
[14:04] <JEEB> Keshl, it actually doesn't get better than x264 in things unless you really push it to the max though :P
[14:05] <Keshl> relaxed: http://pastebin.com/Z53PHyXT
[14:05] <JEEB> it gets very slow then, but that's what you get (about ~10 times slower than x264's placebo)
[14:05] <Keshl> JEEB: Must be pushed to max by default, then. That'd explain why it takes 20 minutes with 264 and about 4 hours with 265. XD
[14:05] <JEEB> no, it is not
[14:05] <JEEB> and yes, it is generally slower
[14:05] <JEEB> also any of the other presets are comparably useless
[14:05] <JEEB> just max it out and you are going to get some efficiency :P
[14:06] <relaxed> Keshl: "built on Jan 22 2014"
[14:06] <Keshl> relaxed: Too old.
[14:06] <Keshl> relaxed: Support for x265 didn't come out until Feb 11th.
[14:06] <relaxed> The freaking build on the download page is "built on Feb 25 2014"
[14:06] <Keshl> Oh.
[14:06] <Keshl> wait, wh -- ..
[14:06] <Keshl> .....
[14:06] Action: Keshl facedesks.
[14:06] <relaxed> the git build
[14:07] <Keshl> Hold, misclicked.. x.x
[14:07] <Keshl> yep, I see it now, sorry, thanks x.x
[14:07] <JEEB> but yeah, for whatever reason libx265's preset placebo doesn't actually do the same as x264's.
[14:07] <JEEB> so you don't get ref16 bframes16
[14:07] <JEEB> which does help
[14:08] <JEEB> and then of course you need to set tune to ssim
[14:08] <JEEB> because otherwise you get BLUUURR (PSNR optimization)
[14:09] <Keshl> D: It hates me!
[14:09] <Keshl> http://pastebin.com/MGV1Jn5B -- Says that the aspect ratio's zero. I don't understand how he's getting that idea o_o
[14:10] <Keshl> JEEB: I already messed it up horribly once. XD
[14:10] <Keshl> That was *fun* =D!
[14:10] <JEEB> anyways, just to make sure you know, x265's development is purely done by MCW, which has zero of the x264 devs on board
[14:10] <JEEB> they just got the name :P
[14:11] <Keshl> Really though. I already get 50% filesize reductions with the same visual quality using just -vcodec libx265 -x265-params crf=22 -preset medium   and nothing else. It's nuts. o_o
[14:11] <JEEB> (and the rights to use the x264 code base)
[14:11] <relaxed> and that sucks :(
[14:11] <JEEB> Keshl, no
[14:11] <JEEB> no
[14:11] <Keshl> No?
[14:11] <JEEB> it does not produce that
[14:11] <JEEB> also don't mux into matroska with ffmpeg
[14:11] <JEEB> only output raw hevc files
[14:12] <JEEB> and mux with L-SMASH or so
[14:12] <Keshl> I don't know what you're saying D:
[14:12] <JEEB> Output #0, matroska, to 'OU265.mkv':
[14:12] <Keshl> Here, gimme a second, lemme just make this dead simple..
[14:12] <JEEB> THIS IS NOT GONNA BECOME SPEC-COMPLIANT
[14:12] <relaxed> I may start including l-smash in my archive
[14:13] <JEEB> because libavformat doesn't support HEVC muxing into anything yet (except maybe, maybe in to mpeg-ts, but I'm not sure about that either)
[14:13] <JEEB> as far as containers go
[14:13] <JEEB> so just stick to dot-hevc
[14:13] <relaxed> what are the x264 looking at?  kvazaar?
[14:13] <Keshl> Oh, you're talking specs. I thought you were gunna say that it doesn't encode, period.
[14:13] <relaxed> x264 devs*
[14:13] <JEEB> oh it will encode and mux shit into matroska
[14:14] <JEEB> but it's not standard
[14:14] <Keshl> I was just about to toss you an FTP to an .mkv with 265.
[14:14] <Keshl> oÉo.
[14:14] <JEEB> and it will ONLY work with lavf
[14:14] <Keshl> ... This explains much. oÉo.
[14:14] <JEEB> it only "works" because lavf has a default muxing mode for matroska etc.
[14:14] <JEEB> which is just "throw extradata there, and actual data there, and give it some random ID specified as a FourCC"
[14:15] <Keshl> So what's the file extension? .hevc?
[14:15] <JEEB> yes
[14:15] <JEEB> relaxed, btw if you're going to build L-SMASH you pretty much have to patch it with https://github.com/l-smash/l-smash/pull/14
[14:15] <JEEB> because the cli app has HEVC muxing disabled right now
[14:15] <Keshl> Anything else, oÉo? Besides the -placebo stuff. I know it'll help, but right now I'm still trying to figure out how low I can push the CRF before things get too big.
[14:15] <JEEB> I don't usually go lower than 24
[14:15] Action: Keshl has no clue what L-SMASH is. D:
[14:16] <JEEB> crf 24, preset to placebo, refs and bframes to 16, tune to ssim
[14:16] <JEEB> that's pretty much sanely'ishly maxing it out
[14:16] Action: Keshl is obsessed with quality and fine detail. With H.264, he woulfn't go higher than 18.
[14:16] <JEEB> there's the full ME mode
[14:16] <JEEB> Keshl, well you can't really do visual comparisons with too high rate now can you?
[14:16] <JEEB> because there will be no differences :P
[14:16] <Keshl> That's the point, oÉo.
[14:17] <Keshl> At 18, as long as you don't pause, the encoded video looks the same as the raw, uncompressed, freshly-captured video.
[14:17] <JEEB> anyways, the crf range is completely different from x264
[14:17] <JEEB> personally at least with 720p 24 was as low as I'd go for testing purposes
[14:18] <Keshl> I know it's different, that's why I'm uisng medium and xperiemtning with crf now, rather than placebo and trying to experiement with crf. XD
[14:18] <Keshl> ... Whoa, typos.
[14:20] <JEEB> anyways, always use tune ssim with x265 until they default to it or visual psy-opts
[14:21] <JEEB> otherwise AQ and other stuff will just be disabled
[14:21] <Keshl> Whus AQ, oÉo?
[14:21] <JEEB> adaptive quantization
[14:21] <Keshl> Ohhhh..
[14:21] <Keshl> That's why it always says q=0.0, isn't it? x.x
[14:21] <Keshl> And why parts with -- Ohhh..
[14:22] <Keshl> Where were you twelve hours ago? My CPU would be screaming less.. XD
[14:22] <JEEB> uhh, no. I don't even know if the q value says anything except when you're using constant quantizer mode
[14:22] <Keshl> It does. When you use libx264 it says stuff, oÉo.
[14:22] <Keshl> And it changes depending on the amount of motion in the scene.
[14:22] <JEEB> yes, but that isn't really useful per se
[14:23] <JEEB> and has nothing to do with AQ/no AQ
[14:23] <Keshl> o.O?
[14:24] <Keshl> o.O''' Uh, encoding to .hevc made stuff break *weird*.
[14:24] <Keshl> Now the seek bar doesn't work, the timing information seems to be lost, it's playing at 25 FPS (Because it guessed, not because it's lagging -- CPU and GPU usage are down really far), and the audio's not playing.
[14:24] <JEEB> uhh
[14:25] <JEEB> let's start with the most obvious
[14:25] <JEEB> it's raw HEVC, there's no audio
[14:25] <Keshl> ... Oh.
[14:25] <JEEB> second, I'm pretty sure x265 doesn't yet write any VUI timing information because there are patches for that on its mailing list
[14:25] <JEEB> third, it's raw HEVC, no real chances of seeking in that without indexing it before hand
[14:25] <Keshl> What do you suggest I do, then? D: Audio's kinda important. So's playing at 60 FPS. D:
[14:26] <JEEB> mux it with L-SMASH patched with the patch I have in the pull request?
[14:26] <Keshl> I don't know what L-SMASH is. o_o
[14:26] <JEEB> muxer -i hurr.hevc?fps=60 -o out.mp4
[14:26] Action: Keshl hugs the word "hurr" -É-..
[14:28] <JEEB> but yeah, clone the main L-SMASH repo and then grab that one commit from my repo on the top :P
[14:28] <JEEB> and L-SMASH itself is very simple to compile, ./configure and then make
[14:29] <Keshl> Any deps? I'm on a weird-butt version of Gentoo that has like, zero usable developement libraries in its repo. x.x
[14:29] <JEEB> not really
[14:29] <Keshl> And earlier, you said to use -ref 16. It claims it's not found. Did you miss a character, oÉo?
[14:29] <JEEB> it's either refs or ref
[14:29] <JEEB> same for bframe(s) 16
[14:29] <Keshl> Refs, oÉo.
[14:29] <Keshl> ...Now bframes isn't found. XD
[14:29] Action: Keshl puts singular.
[14:30] <Keshl> Also not found!
[14:30] <JEEB> also I'm pretty sure you have to use x265opts for those
[14:30] <Keshl> Maybe it wants cframes. oÉo.
[14:30] <JEEB> because daemon404 decided that the way that shit was done for libx264 was crazy
[14:30] <relaxed> indeed
[14:30] <Keshl> x265-params got it, oÉo.
[14:30] <Keshl> Wait. Nope.
[14:30] <Keshl> ... Wait, that's me being an idiot.
[14:31] <Keshl> Yeah, got it. -É-.
[14:31] <Keshl> Forgot that they're seporated by commas in lists like that, not spaces.
[14:31] <Keshl> OH GOSH THE PAIN °²°
[14:31] <relaxed> I'll proabaly include a static x265 with my build along with kvazaar.
[14:31] <JEEB> kvazaar's pretty cool
[14:32] <JEEB> and much simpler to participate in
[14:32] <Keshl> One frame encoded, roughly every 15, sixteen, seventeen..
[14:32] <Keshl> 20ish seconds. o_o
[14:33] <Keshl> Whelp, that's why I wanted the Linux build. Least I have a second computer to run other STUFF with.
[14:33] <Keshl> Now, L-thingy.. Okay, lesse how bad I mess this up.
[14:33] Action: Keshl cracks his compiling knuckles.
[14:33] <relaxed> JEEB: think your patch will be included soon?
[14:33] Action: Keshl realizes he only knows how to use Mercurial and SVN. x.x
[14:35] <Keshl> I officially have no idea what I'm doiiiiing o_o
[14:35] <Keshl> That statement probably shoudl've been said when I first mentioned twosixfiiiive o_o
[14:42] <JEEB> relaxed, VFR Maniac/muken's official opinion is that it will fully go in only after the FDIS ballot ends
[14:42] <JEEB> which is in two months
[14:42] <JEEB> technically you can't really have changes after the FDIS ballot has been started, but he is kind of heavy on the "only official specs" stance
[14:44] <JEEB> so yes, for the time being you will have to get the stuff from that pull request
[14:44] <JEEB> I'll try to get some parts of it to be merged before that, but I bet the cli won't be enabled before the FDIS ballot is over
[14:54] <Keshl> Okay, stuff hates me. x.x Firstly, let me say that I don't plan on erasing the original uncompressed file, and I only plan to send these videos to one person because he's got a really lousy connection that likes to randomly drop to 10kb/sec. I know I'm breaking standards here, but if it plays on his system, that's all I care about at the moment, otherwise I'm still using H.264 for the foreseeable future primarily due to support for
[14:54] <Keshl> hardware acceleration.
[14:55] <Keshl> My windows system has a processor half as fast as my Linux system. On windows, I'm running "./ffmpeg -ss 135 -i MushVis_RAW.flv -vcodec libx265 -x265-params crf=22,refs=16,bframes=16 -preset placebo -movflags faststart -pix_fmt yuv420p -tune ssim -vf scale=1366:768 -acodec copy -t 20 out16.mkv". It works fine. Encoding's still in progress.
[14:55] <JEEB> just make sure that file only gets played with lavf and that it doesn't go around the internets
[14:56] <Keshl> That's ensured, don't worry. oÉo.
[14:56] <Keshl> On Linux, that very same command results in an error saying a value of "22,refs=16,bframes=16" is invalid for "crf", this I have to specify -x265-params three times. That makes it take. .. Almost.
[14:56] <Keshl> Even if I take out the refs and bframes, I get this error on the Linux system only: "Sample Aspect Ratio width must be greater than 0".
[14:56] <Keshl> Any suggestions? xwx
[14:58] <JEEB> btw, you will be using AVC/H.264 for the foreseeable future because it will just be bringing you more bang for the buck for right now, maybe in a 6-12 months period it will get better
[14:58] <JEEB> granted, the fact that you can kind of get better results with x265 already is kind of promising given how damn broken that thing can be with random revisions
[14:58] <Keshl> What do you mean "bang for your buck"? If you mean in terms of filesize vs quality, no. 265, at least for the videos I'm using, is already 50% smaller with the same visual quality, even without refs= and bframes= specified.
[14:58] <Keshl> And without -tune.
[14:59] <JEEB> then you're testing in a bit rate range that is useless for testing :P
[14:59] <Keshl> Here, seriously, lemme send you a 20 second clip. No refs, no bframes, no tune, and preset set to medium, not placebo.
[14:59] <Keshl> o.O? What do you mean by that, exactly?
[14:59] <JEEB> and/or your content can take a lot of blurring without visual problems
[14:59] <Keshl> ... No, trust me it can't. XD
[15:00] <JEEB> then it's the first case
[15:00] <JEEB> and I'm not going to lecture you on how to do testing properly because I will be a nerve wreck rather quickly
[15:01] <Keshl> oÉo.. I guess 18,000 could be considered a bit high <É<'
[15:01] <JEEB> in any case, if you are CURRENTLY getting "50% smaller and looks better or no difference seen", then you're way higher than you should be in the bit rate range
[15:01] <Keshl> 140,000 could be considered even higher. o_o Already made that mitake. x.x
[15:01] <Keshl> ... Wait. I think I severely misunderstand wgat a bitrate is.
[15:01] <JEEB> because x265 most definitely is NOT better than x264 on any standard preset
[15:01] <Keshl> *what
[15:02] <nyuszika7h> hi, does anyone know how to burn in subtitles from an .srt file?
[15:02] <Paranoialmaniac> btw i dont recomment use of ref=16. this has doubt of a bug of the spec...
[15:02] <Keshl> I've been assuming that bitrate is how fast data is read from storage medium. Is that what it is?
[15:02] <JEEB> it is the amount of bits in average used over the clip for a second of content
[15:03] <Keshl> "used" how? By reading from storage medium or by decompressing and tossing at the screen?
[15:03] <JEEB> oh for fuck's sake
[15:03] <JEEB> THE FUCKING BIT STREAM YOU IMBECILE
[15:03] <JEEB> or byte stream if you do Annex B
[15:04] <JEEB> coded picture sizes
[15:04] <Keshl> I really don't know what that is. o_o Look, you've been doing this for years and years. I've been doing this, maybe six or seven months. You program for the project. You obviously have a deep understanding of this. I don't.
[15:04] <Keshl> What you think is common knowledge, is not common knowledge.
[15:05] <Keshl> I am trying my best here. I am admitting I may be wrong about stuff rather than just assuming that I'm right and there's a bug somewhere.
[15:05] <Keshl> Would you rather have someone like me, or someone saying "rawr ffmpeg sux, fix it!!1"?
[15:05] <JEEB> I would ignore the latter, so I probably would rather have him
[15:05] <Keshl> Interesting train of thought..
[15:06] <JEEB> in this case you're under my limit of ignoring someone, and thus I end up using my time
[15:06] <JEEB> in any case, if you don't understand the concept of average bit rate over the whole clip you're really out of the range I can help, sorry :P
[15:07] <Keshl> And this is one of the many reasons you'll never see me ask for commit access here. XD
[15:07] <JEEB> because it's literally <amount of bits the whole clip takes> divided on the length of the clip in seconds
[15:08] <Keshl> I think we're stuck in an infinite loop. Shall we move on? XD
[15:11] <Keshl> Regardless, I do appreciate the help you give and you not exploding on me until now.
[15:11] Action: Keshl snugs at -É-
[15:45] <esperegu> Keshl: fuck me. probably was a mic somewhere that was causing this.
[15:46] <esperegu> Keshl: thx 4 your help
[15:46] <Keshl> Welcomes, oÉo.
[18:06] <FearlessPidgin> Is there any way to determine the deinterlace method needed, based on the frame types obtained from ffprobe?
[18:07] <JEEBsv> no
[18:08] <JEEBsv> since you can't really decide such things by that only
[18:08] <JEEBsv> but in any case if you are just running with eyes closed and want to automatize you either say no to interlaced content from whomever is providing stuff to you, or you yadif all pictures noted as interlaced.
[18:09] <JEEBsv> because that's the only way you're getting something semi-sane automated
[18:13] <FearlessPidgin> JEEBsv: But since there are different types of interlacing (Inverse Telecine comes to mind), does yadif handle a wide range well?
[18:17] <klaxa> my test encodes i did some months ago with yadif and telecined source weren't so good
[18:17] <klaxa> note: months ago
[18:17] <JEEBsv> it's not perfect, esp. if you default to full bobbing
[18:18] <JEEBsv> but I know many people being completely OK with using normal yadif
[18:18] <klaxa> back then i used fieldmatch,decimate which seemed to work better
[18:18] <JEEBsv> for IVTC'd content
[18:18] <JEEBsv> and heck
[18:18] <JEEBsv> you can't automate IVTC
[18:18] <JEEBsv> s/IVTC'd/telecined/
[18:18] <JEEBsv> anyways, the bottom line is that if you don't know what you're being fed
[18:19] <JEEBsv> you can only automate stuff with yadif, pretty much
[18:19] <FearlessPidgin> the problem is I don'y know what type it is, that's why I wanted to find out based on frame types so I could choose the correct deinterlace method
[18:19] <JEEBsv> well the frame types don't contain anything related to that
[18:19] <JEEBsv> just the coding mode
[18:19] <klaxa> sounds like an AI problem to me
[18:20] <FearlessPidgin> Oh... I thought the order of the interlaced frames mattered
[18:20] <FearlessPidgin> klaxa: AI?
[18:20] <klaxa> artificial intelligence
[18:20] <JEEBsv> well, with hard telecine all you will be getting is that all pictures are coded interlaced
[18:20] <JEEBsv> gather much from this other than 'shit's interlaced'
[18:20] <JEEBsv> :P
[18:23] <klaxa> well what one *could* try is to extract a couple of frames and try to find a pattern, as in 3 interlaced frames, 2 full frames
[18:23] <klaxa> that would mean you have to judge whether or not a frame is a full frame based on its actual content
[18:23] <klaxa> which could become hard depending on the content
[18:23] <klaxa> however, the pattern would become easier visible with a lot of frames
[18:24] <JEEBsv> well, IIRC you could mish-mash the fieldmatch filter with yadif somehow, but generally trying to automate with more complexity is very error-prone
[18:24] <FearlessPidgin> yadif has to do it somehow so it can correctly deinterlace right?
[18:25] <JEEBsv> no
[18:28] <FearlessPidgin> Well I don't really need to automate, I just want to be sure I know what type of interlacing I'm dealing with so I can properly get it to progressive with decent quality... Is there any way to do with with some level of certainty?
[18:28] <JEEBsv> using your eyes, mostly
[18:28] <JEEBsv> only way of being relatively sure
[18:33] <FearlessPidgin> JEEBsv: By extracting the frames and looking for patterns of interlaced frames?
[18:35] <JEEBsv> well, fields to be exact, but yes. And patterns of two-fields-per-picture and one-field-per-picture, and possible duplicate pictures
[18:37] <FearlessPidgin> JEEBsv: After that can yadif take input to specify a more specific interlace type?
[18:38] <JEEBsv> well, after you have already checked the content you can then already make the decision on which type of 'creating progressive content out of whatever you have on hand' you need to use
[18:38] <JEEBsv> or multiple ones, depending on if it's mixed content
[18:38] <JEEBsv> (always fun)
[18:38] <FearlessPidgin> oh god
[18:40] <FearlessPidgin> Next question then, how to dump all frames, or a large amount of frames, to images?
[18:41] <JEEBsv> personally if I were to want to do things properly I'd look into frame servers instead of ffmpeg :P
[18:41] <JEEBsv> or something that uses frame-exact methods and lets you check the content, and do stuff like SeparateFields() easily
[18:42] <FearlessPidgin> JEEBsv: Any examples of what your talking about?
[18:42] <JEEBsv> avisynth or vapoursynth
[18:46] <FearlessPidgin> JEEBsv: So you would dump to images with avisynth?
[18:47] <JEEBsv> no, I would use AvsPmod to scroll through the clip
[18:47] <JEEBsv> and look at various parts
[18:49] <FearlessPidgin> "look for various parts"?
[18:49] <JEEBsv> at
[18:49] <JEEBsv> as in, go through the clip's various parts and see what kind of content there is
[18:49] <FearlessPidgin> ah
[18:50] <JEEBsv> (telecined, true interlaced, etc.)
[18:50] <JEEBsv> and when I've gotten enough spots from it done, I then proceed to handling the source correctly
[18:50] <FearlessPidgin> JEEBsv: Would you then yadif after you found out, or use something built into avisynth to do the deinterlacing and feed into ffmpeg?
[18:51] <JEEBsv> yeah, I would then proceed to using TIVTC or QTGMC or whatever is the best tool for the job in avisynth. And then that I would pipe with avs2pipemod or something into ffmpeg or whatever
[18:53] <FearlessPidgin> JEEBsv: I really appreciate the help
[18:55] <FearlessPidgin> JEEBsv: What do you think about DGIndex?
[18:55] <FearlessPidgin> I think that does a sort of frame analysis ?
[18:55] <JEEBsv> DGindex or the d2vsource thingamajig can be used for MPEG-2 input reading
[18:55] <JEEBsv> it doesn't do any ANALYSIS
[18:56] <JEEBsv> as in, worthwhile analysis for your problem :P
[18:56] <JEEBsv> the analyzer will be _you_
[18:56] <FearlessPidgin> ah ok
[18:56] <JEEBsv> looking at the pictures (fields, frames)
[18:57] <FearlessPidgin> since I'm on windows none of this is a problem, but I feel like the Linux guys are left behind when it comes to weird interlacing
[18:58] <JEEBsv> avisynth has worked on wine for over 10 years now
[18:59] <JEEBsv> so they're really not
[18:59] <JEEBsv> and if they want something similar there's always the reimagining called vapoursynth
[18:59] <JEEBsv> which is cross-platform
[20:48] <mikeflynn> I'm looking for a little ffmpeg help if anyone can lend a hand. I just posted this stackoverflow question, but I'm happy to rehash the details in here if that's preferred: http://stackoverflow.com/questions/22024376/ffmpeg-2nd-pass-segfault
[21:02] <llogan> mikeflynn: SO is for programming questions. next time FFmpeg cli tool questions can be asked at Super User (or ffmpeg-user mailing list)
[21:03] <llogan> also your complete console output is required.
[21:03] <mikeflynn> Good point.
[21:03] <mikeflynn> Ok, one second.
[21:03] <llogan> why are you not simply using a libx264 encoding preset?
[21:04] <mikeflynn> I didn't start from scratch, I'm trying to recreate an existing and working transcoding server to a new enviroment so I don't have any of the original context for these decisions.
[21:06] <llogan> why 2-pass? are you trying to get a certain output file size?
[21:07] <mikeflynn> 2-pass is what the current codebase calls. I was hoping to get this up and running without having to do any major code surgery, but at this point I'm open to anything.
[21:09] <llogan> start with something like this: ffmpeg -i input -vcodec libx264 -preset slower -crf 23 -acodec libfaac -q:a 100 output.mp4
[21:09] <llogan> the "slower" preset is probably closest to your previous, messy settings.
[21:11] <mikeflynn> That command in place of the 2nd pass, or replacing both passes?
[21:11] <mikeflynn> Here's the full output of the failing 2nd pass: http://pastebin.com/1Kuc7W5F
[21:13] <llogan> i'm assuming you don't need two passes, so you can just use one pass using -crf.
[21:14] <llogan> https://trac.ffmpeg.org/wiki/x264EncodingGuide
[21:14] <llogan> https://trac.ffmpeg.org/wiki/AACEncodingGuide
[21:14] <llogan> use the highest crf value that still give an acceptable quality. use the slowest preset you have patience for. use those settings for the rest of your videos.
[21:15] <mikeflynn> Yeah, I was going through that last night. I'm just wary to modify this setup too much, as crazy as it might be. This confirmation helps though, I'll dive in for surgery!
[21:15] <mikeflynn> Ok
[21:15] <llogan> but your input is already H.264 and AAC, so just don't re-encode.
[21:16] <mikeflynn> This one was h.264/aac but will have other sources: flv, avi, etc..
[21:16] <mikeflynn> (video upload site)
[21:18] <llogan> what is the output file going to be used for?
[21:18] <elf__> hi, I have a question. I have a video file,  I am trying to extract the audio to a wav file using " -acodec pcm_s16le -ac 2 -f wav output.wav". But if I change output.wav to be just - (-acodec pcm_s16le -ac 2 -f wav - | hexdump), there is a difference. The header of the  wav file has a chunk size. The output.wav has a valid value but the output to the command line always has zero. Has anyone seen this before and is there a known work around?
[21:21] <mikeflynn> @llogan - Web display through a flash or html5 player.
[21:22] <llogan> add "-movflags +faststart" and "-pix_fmt yuv420p"
[21:28] <mikeflynn> Great, thanks. Again, I was nervous to change what they had too much, but this conversation gives me a little more confidence.
[21:28] <mikeflynn> I appreciate the help!
[00:00] --- Wed Feb 26 2014