[Ffmpeg-devel-irc] ffmpeg.log.20170322

[burek]

[00:07:38 CET] <ritsuka> you shouldn't ignore the edit lists btw
[00:56:46 CET] <alexpigment> ritsuka: yeah i was wondering that myself. i'd guess that the old version ignored the edit lists and the new version respected them. in which case, it's probably best to use a newer copy of FFMPEG and *not* ignore them
[00:58:48 CET] <alexpigment> unless, of course, there's an A/V sync issue caused by the edit lists (this used to happen back in the day, I recall)
[01:10:44 CET] <ZeroWalker> okay was able to steram tcp, but it seems to be quite bad for low latency compared to how rtp handles it (the video part at least). in rtp it seems to just play the frame as fast as possible, which is what i want (as it's VFR, which doesn't seem to work with a tcp stream, it will buffer and play it at specified timebase).
[01:12:06 CET] <ZeroWalker> so, i would probably want to try rtp with tcp (or something, find it really complicated about all these protocols;P). And then maybe stream the audio separately
[04:05:54 CET] <obamoose> hello
[06:07:45 CET] <IRC-Source_62632> Hey, anyone online?
[06:26:07 CET] <IRC-Source_62632> I have a question, can ffserver stream audio in PCM raw format?
[09:04:38 CET] <BridgeUK> Hi, please can someone tell me what a malformed aac bitstream is and how to fix it, thank you very much
[09:15:19 CET] <Threads> -bsf:a aac_adtstoasc
[09:15:27 CET] <Threads> should help
[09:22:10 CET] <BridgeUK> thanks for the reply Threads, when I do "ffmpeg.exe file.mp4 -bsf:a aac_adtstoasc" I get two error messages, "Trailing options were found on the commandline." and "Output file #0 does not contain any stream" please can you tell me what I need to do next
[09:24:54 CET] <BridgeUK> http://pastebin.com/vADjycMa is the full printout
[09:26:28 CET] <BridgeUK> sorry i'm just being a bit overcautious, i don't want to so something wrong which may break the file :)
[09:36:21 CET] <Threads> are you copying or extracting ?
[09:36:26 CET] <Threads> what do you need fully
[09:44:24 CET] <BridgeUK> Threads, what it is is i downloaded a couple videos with youtube-dl and noticed youtube-dl said "malformed aac bitstream" and i wasn't sure what that meant apart from it was an ffmpeg message
[09:45:21 CET] <BridgeUK> the audio is fine on the files, it's just i wasn't sure what the malformed aac bitstream meant
[10:25:07 CET] <Hawoo_> Hi:) I am a student from GSoC program. I would like to participate in VMAF video filter program but the specified qualification task was already completed, I wonder is there any alternative task I can choose? Thanks a lot!
[10:25:45 CET] <durandal_170> Hawoo_: wrong channel
[10:26:10 CET] <durandal_170> Hawoo_: #ffmpeg-devel
[10:27:22 CET] <Hawoo_> So sorry
[10:51:38 CET] <BOFH_org> Hi all
[10:54:58 CET] <BOFH_org> I try to recode a m2p file to a h264 avi. the stream is greyscale and although there is an audiotrack in the m2p, it's just static. So I'd like to drop the audio track when recoding. I use the following commandline: ffmpeg -i input.m2p -q:v 0 -pix_fmt grey -vcodec h264 -an output.avi. But now I get repeating warning like this: Past duration 0.999992 too large. This warning doesn't appear when I exclude '-an' from the commandline. What's going on here? Can
[10:54:59 CET] <BOFH_org> I just ignore these warnings?
[10:56:27 CET] <nyuszika7h> Those warnings can be usually ignored
[10:56:52 CET] <nyuszika7h> Also why would you put H.264 video in AVI?
[10:56:57 CET] <nyuszika7h> You can, it's just illogical
[10:57:18 CET] <nyuszika7h> AVI generally stores XviD, for H.264 use MP4/MKV
[10:57:21 CET] <vlt> BOFH_org: Do those warning appear when you use a decent container format?
[10:57:42 CET] <nyuszika7h> I get such warnings with MKV sometimes
[10:57:50 CET] <nyuszika7h> But the output file is fine
[10:57:51 CET] <BOFH_org> let me try.
[10:58:42 CET] <BOFH_org> exactly the same error, regardless the container I use. the warnings disappear when removing the '-an' flag though
[11:28:30 CET] <panpeter> Hi all, is "gst_debug_bin_to_dot_data" in all gstreamer1.0 versions available?
[11:59:51 CET] <BOFH_org> ahh. it seems the "past duration too large" errors disappear when using .mp4
[12:00:45 CET] <BOFH_org> nyuszika7h: indeed, it seems the warnings can be ignored. video output seems fine. thanks
[12:11:11 CET] <ZeroWalker> is it possible to output CFR with VFR input (in code). I set the framerate at the encoder, and input is VFR and just timestamp based live. but the output is the same as the input. Which is fine usually, but it seems to be a bit problematic when trying to stream as it wants CFR, unless i am missing something
[12:11:16 CET] <panpeter> Ok it seems not to be the case, in 1.8.3 "gst_debug_bin_to_dot_data" is available in 1.2.4 only "gst_debug_bin_to_dot_file" and "gst_debug_bin_to_dot_file_with_ts" is available
[12:16:56 CET] <panpeter> Is there a way to set "GST_DEBUG_DUMP_DOT_DIR" directly for gstreamer instead of using an environment variable?
[12:54:54 CET] <ZexaronS> So MPEG4 isn't really a successor to MPEG2 - what about a real MPEG2 successor
[12:55:30 CET] <ZexaronS> Online streaming is definitely another class than MPEG2 was in
[12:57:01 CET] <ZexaronS> A compression that compresses more but also loses more information is not considered "better", weird monikers, it's really more specific, better in filesize but not better overall
[13:03:11 CET] <flux> zexarons, so you're looking for better quality? in the end it all comes down to required bitrate and achieved quality at that bitrate. for that function I believe H264 and H265 handily beat/succeed MPEG2.
[13:04:03 CET] <ZexaronS> the hardcore guys around MPC-HC always say MPEG2 is better looking
[13:04:16 CET] <ZexaronS> but at high bitrates imo only
[13:04:19 CET] <kerio> the hardcore guys are stupid
[13:04:21 CET] <ritsuka> -.-
[13:04:32 CET] <kerio> like that's not true in any way shape or form
[13:04:44 CET] <ZexaronS> this is something MP4 and HEVC can't do
[13:04:50 CET] <ZexaronS> that's what I heard
[13:05:07 CET] <kerio> h264 can be literally lossless
[13:05:25 CET] <kerio> so i have absolutely no idea what those guys are smoking
[13:06:13 CET] <ZexaronS> But I have noticed the same thing with AVC it's always so blocky compared to silk smoothness of MPEG2
[13:06:23 CET] <ZexaronS> but im not an expert maybe it wasn't a good comparison
[13:06:24 CET] <kerio> at the same bitrate?
[13:06:32 CET] <ZexaronS> sorry i don't remember
[13:06:44 CET] <kerio> then what is the point
[13:07:11 CET] <nido> does doom9 still do its comparisons?
[13:07:24 CET] <flux> could even be interlaced/progressive mixing up
[13:07:39 CET] <flux> interlaced sure was popular with mpeg2.
[13:07:46 CET] <ZexaronS> The point is, when they designed HEVC and AVC they focused more on getting the size down than expanding quality features and designing it in quality in mind, the compression probably isn't meant to preserve quality
[13:07:58 CET] <nido> hrm... doom9 doenst work anymore
[13:08:10 CET] <kerio> http://streaminglearningcenter.com/images/2mbps_horse.jpg
[13:08:18 CET] <kerio> ZexaronS: no they didn't
[13:08:22 CET] <kerio> and yes it is
[13:08:28 CET] <kerio> that is literally the whole point of lossy compression
[13:11:19 CET] <ZexaronS> well i must have heard something a long while ago like 2 years
[13:11:27 CET] <ZexaronS> but there was also discussion of color
[13:11:48 CET] <ZexaronS> I had some old dvd archival remastering to do a year back
[13:12:15 CET] <ZexaronS> I poked into all things transcoding for the first time for serious, but i had some early beginner experience
[13:12:23 CET] <ZexaronS> but im still not that of an expert
[13:13:29 CET] <ZexaronS> when HDR and WCG arrive, would these codecs be fine? they weren't designed around keeping good color info ?
[13:15:10 CET] <ritsuka> yes, there are already using h.264/h.265 for hdr
[13:15:54 CET] <ZexaronS> I'm trying to find that picture
[13:40:17 CET] <ZexaronS> oh ops
[13:40:30 CET] <ZexaronS> that was just how MadVR renders compared to EVR in MPCHC
[13:40:52 CET] <ZexaronS> MadVR looks more brown while EVR more reddish, but these could just be my settings at the time who the heck knows
[14:23:43 CET] <jonascj> Is it possible to specify a specific lame library version in ffmpeg, or will ffmpeg always use what is installed in predefined dirs e.g. /usr/lib ?
[14:24:25 CET] <jonascj> by specifying I mean if you could point it to /tmp/libmp3lame.so.0
[14:36:59 CET] <kepstin> jonascj: on linux? you can use standard linker mechanisms e.g. LD_LIBEARY_PATH or LD_PRELOAD to affect which library is used.
[14:39:02 CET] <jonascj> kepstin: on linux yes, I've never played much with LD_LIBRARY and lD_PRELOAD, maybe now is the time. I'll look into it. Otherwise I would jsut relink the /usr/bin/ file to the version I wanted stored elsewhere.
[14:39:50 CET] <kepstin> jonascj: simplest way to do it is something like `LD_PRELOAD=/tmp/libmp3lame.so.0 ffmpeg ...`
[14:40:53 CET] <jonascj> thanks, I'll try that.
[14:42:41 CET] <kepstin> why are you switching out the lame encoder anyways? my impression is that lame hasn't really changed in quite a few years...
[14:43:24 CET] <jonascj> kepstin: for kicks, but no, last stable release was 2012 I think
[14:43:31 CET] <jonascj> Just wanted to try and build it myself etc.
[14:53:37 CET] <jonascj> do you know why ffprobe shows "encoder lavc57" instead of something like: "encoder: lame" when I encode audio as "ffmpeg -i file.wav -c:a libmp3lame file.mp3"?
[15:11:04 CET] <BtbN> Because libavcodec57 is the encoder.
[15:11:52 CET] <jonascj> BtbN: which relies on the lame library?
[15:12:04 CET] <BtbN> It relies on whatever you tell it to use
[15:12:06 CET] <jonascj> internally
[15:12:12 CET] <jonascj> alright, thanks
[15:18:13 CET] <alexpigment> BtbN: i saw you had a commit for removing the Nvidia NVENC 'default' preset because there were no quality benefits. do you have any info about what sets it apart from other presets? is it possible it has any advantages on older (or newer) hardware?
[15:18:22 CET] <thebombzen> jonascj: because it was encoded with libavcodec, it uses that. it turns out libavcodec just delegates the actual encoding to libmp3lame, but that's abstracted away from libavformat and the tohers
[15:18:44 CET] <BtbN> alexpigment, "what sets it"?
[15:18:52 CET] <BtbN> I'm not aware of any point for it
[15:19:03 CET] <thebombzen> what does the -preset 0 even do with nvenc
[15:19:24 CET] <alexpigment> "what sets it apart" = "what makes it unique"
[15:19:28 CET] <thebombzen> setting the preset to "default" doesn't actually change anything
[15:19:29 CET] <BtbN> if you want high quality, you go hq, if you want high performance, you go hp. low latency, there are ll presets.
[15:19:54 CET] <BtbN> The only unique quality of it is that it's confusing and seemingly pointless
[15:20:05 CET] <thebombzen> also what's the difference between the various things like "medium" and "hq"
[15:20:06 CET] <alexpigment> a very scientific answer there ;)
[15:20:16 CET] <thebombzen> medium says "hq 1 pass"
[15:20:28 CET] <alexpigment> well, i know in terms of speed, there are ver little *speed* differences between a lot of the presets
[15:20:35 CET] <alexpigment> *very
[15:20:47 CET] <BtbN> fast/medium/slow are convenience presets made by ffmpeg, that set a few other options to mimic x264 presets
[15:20:58 CET] <BtbN> There are huge speed differences betweem them
[15:21:11 CET] <alexpigment> yeah, i'm specifically talking about NVENC presets
[15:21:15 CET] <BtbN> hp can easily do several 1000 fps, while hq with 2 pass(i.e. slow) barely handles 100 or 200
[15:22:23 CET] <alexpigment> anyway, i wanted to make sure I could just safely ignore "Default". but I was really wondering if there was some attribute about default that isn't useful for one application but may be useful for another
[15:22:48 CET] <BtbN> It's not well defined in any way, and I'm actually thinking about deprecating it in ffmpeg as well
[15:22:59 CET] <BtbN> Making it an alias for medium
[15:23:27 CET] <alexpigment> fair enough. thanks for the extra info
[15:29:47 CET] <alexpigment> so is there any documentation on the fast/medium/slow ffmpeg presets for NVENC? i notice, again, that there isn't a huge speed difference on my end between fast and medium - only 5%. and just slightly faster than medium, but warns that "The variable bitrate rate-control requires the 'qmin' and/or 'qmax' option set."
[15:30:17 CET] <alexpigment> *and slow is just slightly faster than medium
[15:30:22 CET] <alexpigment> sorry, no coffee yet
[15:32:00 CET] <DHE> alexpigment: not much beyond what NVENC describes which is little more than "quality improves, framerate decreases". most of the docs I've seen are just framerate estimates to give you an idea how much you lose by choosing slow
[15:33:36 CET] <alexpigment> does the warning i mention above imply that slow doesn't have any effect outside of a particular rate control method?
[15:33:43 CET] <alexpigment> e.g. 2pass
[15:34:33 CET] <alexpigment> in other words, when doing a single pass encoding, why do fast and medium give no warning messages, but slow warns about using qmin and qmax?
[15:39:55 CET] <BtbN> alexpigment, if you get that warning, you should update your ffmpeg. It was removed.
[15:44:52 CET] <alexpigment> BtbN, ok confirmed. i was using the last zeranoe build i downloaded last month or so. my own build doesn't give the same error
[15:45:03 CET] <BtbN> it's not an error
[15:45:04 CET] <alexpigment> and now i'm also seeing a more noticeable different in FPS
[15:45:06 CET] <BtbN> just a pointless warning
[15:45:14 CET] <alexpigment> yeah, i meant warning, sorry
[15:45:34 CET] <alexpigment> man, i need to drink coffee. these typos are out of control :)
[15:46:25 CET] <alexpigment> but yeah, now i'm seeing Slow at 400fps, Medium at 560fps, and Fast at 600fps
[15:46:48 CET] <BtbN> which card is that on?
[15:46:55 CET] <alexpigment> 1060
[15:47:28 CET] <BtbN> if you want to truely bench only nvenc, make sure to use something like yuvtestsrc and -f null as output
[15:49:47 CET] <alexpigment> well, i don't need to do any synthetic benchmarking - i'm using a 1080p60 source which is generally what i'll be working with
[15:50:02 CET] <alexpigment> i just wanted to make sure i was seeing the correct differences between the speed-based presets
[15:50:16 CET] <alexpigment> which i now am, thanks to your tip about using a newer ffmpeg build
[15:50:43 CET] <BtbN> if you are going for livestreaming, there is absolutely no point in going for any but the highest quality preset
[15:52:24 CET] <alexpigment> well, not for livestreaming really, but for draft previews. just something quick and dirty to put on the screen. final encodes will use x264, naturally
[15:53:12 CET] <alexpigment> but yes, i can see that the FPS is high enough that realtime is easily possible on the highest quality presets
[16:11:01 CET] <thebombzen> BtbN: is that the slow preset?
[16:14:15 CET] <panpeter> Is there any internal way to set GST_DEBUG_DUMP_DOT_DIR without using an environment variable?
[16:17:21 CET] <thebombzen> why would you need to do that?
[16:37:37 CET] <erbal> Hi, I'm trying to cross-compile ffmpeg to target Android but having issues. Anyone have any experience with that? Or compiling ffmpeg in general?
[17:11:36 CET] <BtbN> thebombzen, depending on the content it's either llhq+2pass or hq+2pass
[18:44:45 CET] <bmduser> can anyone tell me if there is a precompiled binary for windows that supports the blackmagic decklink duo2?
[18:46:50 CET] <bmduser> because i have been trying everything and all i get is a black recording
[18:47:15 CET] <bmduser> here is the command i am using -f dshow -video_size 1920x1080 -pixel_format uyvy422 -framerate 25 -i video="Decklink Video Capture (2)" -codec:v mpeg4 -pix_fmt yuv420p -preset ultrafast -an -crf 0 output.mp4
[18:47:59 CET] <oerg866> why are you setting the pixel format twice
[18:48:10 CET] <bmduser> I am sure that I have a 1080p video source at 25 fps on every single input on the capture card
[18:49:20 CET] <bmduser> oerg866: i am quite new to this and have been trying it differently
[18:59:33 CET] <bmduser> what i am really trying to do is to stream a backmagic capture card's input to VLC
[19:00:51 CET] <bmduser> with this command       ffmpeg.exe -f dshow -video_size 1920x1080 -pixel_format uyvy422 -framerate 25 -i video="Blackmagic WDM Capture (2) " -f mpegts udp://127.0.0.1:5000?pkt_size=1316
[19:01:55 CET] <bmduser> and all i am getting is a black screen so i thought its a case of me not getting the right input settings on the blackmagic card.
[19:02:17 CET] <bmduser> streaming an mp4 file to vlc works though
[19:42:57 CET] <faLUCE> Hello. If I decode a MJPEG packet to AV_PIX_FMT_YUYV422 format, the resulting AVFrame is filled in data[0], data[1] and data[2].... why that? AV_PIX_FMT_YUYV422 is packed, not planar
[20:51:53 CET] <matkatmusic> off topic: anyone know what are good bitrate settings for exporting?   I've got some raw footage from a camera that says the data rate is 40.72Mbit/sec
[20:53:00 CET] <dystopia_> depends on the type of content
[20:53:30 CET] <dystopia_> and resolution
[20:55:04 CET] <matkatmusic> 1920x1080 both in and rendered
[20:55:09 CET] <matkatmusic> piano tutorials
[20:55:18 CET] <matkatmusic> so, mostly static footage
[20:55:35 CET] <matkatmusic> except for moving hands
[20:55:40 CET] <alexpigment> matkatmusic: is there a reason you aren't able to use CRF?
[20:55:49 CET] <matkatmusic> what is CRF?
[20:55:54 CET] <dystopia_> yeah he can heh
[20:55:58 CET] <matkatmusic> I'm doing the editing in Screenflow
[20:55:59 CET] <dystopia_> so footage is low motion
[20:56:16 CET] <dystopia_> you could use crf 17 or crf 18
[20:56:21 CET] <dystopia_> it will compress well
[20:56:31 CET] <alexpigment> does Screenflow not have a CRF mode?
[20:56:31 CET] <matkatmusic> I've been using 3000kbit/sec because I didn't see any difference when I used higher settings
[20:56:46 CET] <alexpigment> 3000kbps is not bad for very low motion 1080p
[20:56:51 CET] <alexpigment> but i wouldn't go too much lower
[20:57:35 CET] <alexpigment> especially if it's only 1pass VBR
[20:58:54 CET] <alexpigment> i guess your question didnt' get answered up there. CRF = constant rate factor. it basically allows you to choose a desired quality level, then the bitrate is allocated accordingly
[20:59:22 CET] <alexpigment> instead of -b:v 3000k, just use -crf 17 or 18 as dystopia suggested
[20:59:46 CET] <alexpigment> (i tend to use 18 for things where i still retain a master copy in higher quality)
[21:05:47 CET] <furq> if it's osx only then it's probably using the avfoundation h264 encoder
[21:05:57 CET] <furq> which i assume sucks, but i've never used it
[21:06:48 CET] <alexpigment> 2-pass generally makes sucky things suck slightly less
[21:07:12 CET] <alexpigment> i wish Premiere's encoder sucked less :(
[21:10:44 CET] <llogan> alexpigment: install UT video. then you can export it from AME. then re-encode with ffmpeg. or maybe use a frameserver w/ avisynth
[21:11:21 CET] <alexpigment> llogan: i already effectively do that
[21:11:39 CET] <llogan> what method do you use?
[21:11:40 CET] <alexpigment> i export to an intermediate format and then re-encode with ffmpeg
[21:11:52 CET] <furq> llogan: no avisynth on osx
[21:12:07 CET] <alexpigment> furq: i don't think he's talking about OSX
[21:12:18 CET] <llogan> didn't realize it was os x
[21:12:22 CET] <furq> screenflow is osx only
[21:12:25 CET] <furq> so i assume he is
[21:12:34 CET] <alexpigment> furq: you have two people confused
[21:12:38 CET] <furq> oh nvm
[21:13:17 CET] <alexpigment> no worries
[21:13:19 CET] <furq> i don't see why you'd use ut video for that though
[21:13:23 CET] <furq> ffv1 is builtin
[21:13:29 CET] <llogan> when did that happen?
[21:13:30 CET] <furq> wait. premiere
[21:13:44 CET] <furq> i'm doing well today
[21:13:48 CET] <faLUCE> any help about that? I saw the decoding_encoding example: for decoding, AVCodecContext->pix_fmt is not set. Then, how can I know which pixel format is used by the codec context for the output frame?
[21:14:12 CET] <alexpigment> sorry, i was making an aside about how i disliked Premiere's encoder (Mainconcept, i think), which was really an aside since you mentioned that avfoundation's encoder probably sucks
[21:14:28 CET] <llogan> alexpigment: intermediates. the traditional "AME tax".
[21:14:30 CET] <furq> no this is pretty much my fault
[21:14:32 CET] <alexpigment> i think i'm contributing to furq's confusion here
[21:14:43 CET] <furq> anyway yes mainconcept sucks
[21:14:49 CET] <furq> let's all get angry at that instead of at me
[21:15:02 CET] <alexpigment> ;)
[21:15:13 CET] Action: llogan casts magic missle on mainconcept
[21:15:41 CET] <alexpigment> i remember back in the day that mainconcept used to be a big player in the MPEG-2/DVD game
[21:15:58 CET] <alexpigment> and they just got owned in H.264
[21:16:13 CET] <alexpigment> on the other hand, FFMPEG still can't create a decent looking MPEG-2 DVD imho
[21:17:00 CET] <furq> what would you recommend
[21:17:06 CET] <furq> on the off chance i ever need to make a dvd again
[21:17:08 CET] <llogan> i still use it on occassion because I don't care enough...it's just doovday anyway.
[21:17:18 CET] <furq> i think i used hcenc last time i was called upon to do that
[21:17:24 CET] <alexpigment> hcenc is highly regarded
[21:17:28 CET] <furq> but that was like five years ago and it was already surprising that anyone asked
[21:17:39 CET] <alexpigment> i generally just export from Adobe Premiere because I'm already there and it looks pretty decent
[21:18:18 CET] <alexpigment> I still author a decent amount of DVDs for analog > digital conversions, so it's not exactly a "dead" format to me ;)
[21:18:39 CET] <alexpigment> but yeah, FFMPEG... wow. just horrible MPEG-2 quality with any combination of settings i try
[21:18:49 CET] <furq> it was for my flatmate's short film, which he'd already put on vimeo in 1080p
[21:18:56 CET] <furq> he just wanted to have something to hand out
[21:19:14 CET] <furq> i'm not sure which commissioner he expected to meet who didn't have internet
[21:19:29 CET] <matkatmusic> ok, so lemme see what kinds of codecs are available to export with...
[21:19:43 CET] <llogan> furq: maybe for the old people
[21:20:50 CET] <matkatmusic> ok, there is Lossless - ProRes with Alpha where I can customize the frame rate
[21:21:04 CET] <matkatmusic> and then there is Web HD (h264) where i can specify a bunch of paramters
[21:21:07 CET] <matkatmusic> parameters
[21:21:16 CET] <matkatmusic> everything else is like for ipod/iphone or whatnot
[21:21:30 CET] <furq> there's a lossless prores?
[21:21:41 CET] <matkatmusic> Also, there is plain old apple Lossless
[21:22:27 CET] <furq> if that means prores 4444 then that's not lossless
[21:22:45 CET] <matkatmusic> https://s16.postimg.org/me5uxv6at/Screen_Shot_2017_03_22_at_4_22_06_PM.png
[21:22:46 CET] <furq> it has a lossless alpha channel, which seems useful (i am lying)
[21:22:58 CET] <alexpigment> prores is effectively lossless if you're going to re-encode. it's an intermediate format for a reason
[21:23:07 CET] <matkatmusic> https://s11.postimg.org/5cw8uaqlv/Screen_Shot_2017_03_22_at_4_21_58_PM.png
[21:23:07 CET] <furq> it's "visually lossless" according to apple
[21:23:19 CET] <furq> but i would still complain if a $99 application told me it was lossless
[21:23:32 CET] <matkatmusic> that's for "Lossless" setting
[21:24:05 CET] <llogan> probably qtrle
[21:24:08 CET] <matkatmusic> here's the WebHD: https://s18.postimg.org/6o2udqu6x/Screen_Shot_2017_03_22_at_4_23_41_PM.png
[21:24:28 CET] <furq> not even a 2-pass option ;_;
[21:24:31 CET] <matkatmusic> So, the raw footage Data Rate was 40.72Mbit/Sec
[21:24:33 CET] <Threads> isn't everything from apple just 'BS' on a media level ?
[21:24:50 CET] <alexpigment> matkatmusic - are you trying to make these files as small as possible, or is size even a concern?
[21:25:35 CET] <furq> matkatmusic: https://www.macupdate.com/app/mac/24173/x264encoder
[21:25:36 CET] <matkatmusic> size isn't a concern, but upload time is
[21:25:38 CET] <furq> apparently that works with screenflow
[21:26:45 CET] <matkatmusic> Where do you see that?
[21:26:50 CET] <furq> http://telestreamblog.telestream.net/2010/02/apple-h-264-vs-x264-in-screenflow-better-exports/
[21:27:22 CET] <furq> 2010, so hopefully they haven't broken it since then
[21:28:32 CET] <alexpigment> i don't see any CRF settings in those screenshots
[21:28:38 CET] <furq> i assume it's in a different tab
[21:28:49 CET] <alexpigment> unless "Data Rate: Automatic" means CRF (which i doubt)
[21:29:09 CET] <furq> x264 will be much better even if you can't use crf
[21:29:13 CET] <alexpigment> true
[21:29:52 CET] <alexpigment> but it really comes down to what's trying to be achieved here. if a bitrate of 3000kbps gets good results for matkatmusic and the upload time is acceptable, i'm not sure that x264 is going to make a meaningful difference
[21:30:04 CET] <furq> if screenflow works with any quicktime component then there are probably plenty of others you can use
[21:30:51 CET] <alexpigment> other than, of course, knowing that x264 will produce better quality at a given bitrate below saturation
[21:39:58 CET] <Fenrirthviti> Hmm, is NVDefault an actual mode, or just a preset? Does anyone know what settings it translates to if a preset? Having trouble confirming.
[21:40:12 CET] <Fenrirthviti> (for nvenc, of course)
[21:42:50 CET] <matkatmusic> hey, that link is from 2010
[21:42:57 CET] <matkatmusic> check the date on those comments
[21:44:41 CET] <alexpigment> Fenrirthviti: i just talked to BtbN about this earlier today, but couldn't get a good answer
[21:44:52 CET] <alexpigment> according to him, it's not a good setting and is being deprecated
[21:45:15 CET] <alexpigment> but yes, it's a preset
[21:45:49 CET] <alexpigment> the recommendation is to use fast, medium, or slow instead (and preferrably slow)
[21:45:52 CET] <Fenrirthviti> Yeah, I'm just trying to see what it actually translates to. I have a user complaining that the nvenc presets in OBS are not as good of quality as NVDefault, which I find very hard to believe.
[21:46:17 CET] <alexpigment> wish i knew more, but i don't
[21:46:22 CET] <Fenrirthviti> Our higher quality preset is hq (slow)
[21:47:41 CET] <Fenrirthviti> Thanks though, alexpigment.
[21:48:48 CET] <alexpigment> no problem. i do find it funny that i tried to get the exact same information earlier today ;)
[22:10:46 CET] <Fenrirthviti> alexpigment: I'm still digging, if I find anything I'll let you know.
[22:11:57 CET] <Fenrirthviti> everything I'm finding is saying NVDefault is much higher quality than even slow :\
[22:12:00 CET] <alexpigment> yeah cool. i don't need it as much as you, fwiw. i was really just trying to get more details about each preset because they don't seem to make as significant of a difference as x264 presets do
[22:12:12 CET] <alexpigment> really? interesting
[22:12:14 CET] <Fenrirthviti> but nothing is telling me what it's actually doing.
[22:15:10 CET] <alexpigment> yeah i know
[22:15:13 CET] <alexpigment> "It has most of the features enabled except B-frames."
[22:15:24 CET] <alexpigment> that doesn't tell me much ;)
[22:16:48 CET] <seb_> Hello, I want to use ffmpeg to find the same frame in a stream. So i decided  to use -framemd5 but it doesn't work like i want: the same frame give me 2 md5 different. I suppose that's why it work by packet. Is there a command to have de md5 frame by frame from a stream?
[22:17:05 CET] <seb_> Thanks for your help
[22:18:09 CET] <Fenrirthviti> alexpigment: That's the ONLY note I can find about it. I don't even see it referenced in nvidia's docs, though those are lacking in more than one aspect
[22:18:24 CET] <Fenrirthviti> Really annoying.
[22:26:16 CET] <Fenrirthviti> reading the nvenc documentation and it doesn't have anything listed either. it just says "the client can use the presetGUID for configuring the encode
[22:26:16 CET] <Fenrirthviti> session directly. This will automatically set the hardware encoder with appropriate
[22:26:16 CET] <Fenrirthviti> parameters for particular use-case implied by the preset."
[22:30:06 CET] <Fenrirthviti> Just read up on your conversation with BtbN, though, and it looks like there was a misunderstanding. NVDefault is not the same as default
[22:30:13 CET] <alexpigment> ohhhhh
[22:30:23 CET] <alexpigment> well, that explains that
[22:30:29 CET] <alexpigment> is NVDefault not exposed in FFMPEG?
[22:30:32 CET] <alexpigment> just OBS?
[22:30:41 CET] <Fenrirthviti> It's exposed by the nvenc API, yeah
[22:30:58 CET] <Fenrirthviti> so not sure if ffmpeg has an equivalent.
[22:31:16 CET] <alexpigment> i just naturally assumed that the "default" preset was the same as NVDefault
[22:32:06 CET] <Fenrirthviti> Looks like it definitely is not.
[22:32:26 CET] <Fenrirthviti> seems to be some internal API call, but I'm having trouble finding out what it actually does.
[22:32:31 CET] <Fenrirthviti> the API is not very well documented.
[22:33:30 CET] <Fenrirthviti> Or maybe it is the same thing? Man I don't even know anymore :\
[22:33:46 CET] <alexpigment> haha
[22:34:13 CET] <alexpigment> i mean it doesn't help you with that customer, obviously
[22:34:58 CET] <alexpigment> given that i don't use OBS regularly, i presume they still have the NVDefault option right?
[22:35:06 CET] <alexpigment> if so, they should just keep using it and stop complaining ;)
[22:35:42 CET] <Fenrirthviti> We do, it's just changed names.
[22:35:54 CET] <alexpigment> what name is it now?
[22:41:24 CET] <Fenrirthviti> default
[22:41:52 CET] <Fenrirthviti> it was never actually called nvdefault, whoever wrote the original code for the nvenc encoder options in OBS called it nvdefault for some reason.
[22:42:16 CET] <alexpigment> ah, well that explains that discrepancy
[22:42:41 CET] <Fenrirthviti> Sending me on a wild goose chase :\
[22:43:06 CET] <alexpigment> anyway, having dealt with customers about quality issues for quite a few years at this point, i never underestimate the power of screenshots and samples. usually it's like "oh, that's because you're doing something completely different than you said you were"
[22:43:51 CET] <alexpigment> so hopefully you've got actual proof of what they're saying before tearing your hair out just because of what they told you
[22:44:13 CET] <Fenrirthviti> well OBS is free and I'm just community support, I couldn't care less about proving anything to them :P
[22:44:22 CET] <Fenrirthviti> I'm providing the facts, they can do what they want with it.
[22:44:30 CET] <alexpigment> fair enough
[22:44:44 CET] <alexpigment> i'd at least get a screenshot if you haven't already though
[22:44:59 CET] <alexpigment> there's probably another explanation
[22:46:45 CET] <Fenrirthviti> The end user is an idiot who claims shadowplay is better, is the explanation.
[22:54:27 CET] <mac_nibblet> Heya, i'm exporting a bunch of thumbnails from a video using scene detection, I want to figure out when exactly the frame is from, how do I do that ?
[22:54:38 CET] <mac_nibblet> ffmpeg -i video.mp4 -vf "select=gt(scene\,0.6),scale=250:-1" -frames:v 9  -vsync vfr -loglevel verbose images-%d.jpg
[22:54:41 CET] <mac_nibblet> is the command i'm running
[22:54:58 CET] <Fenrirthviti> scene detection what
[22:55:08 CET] <Fenrirthviti> oh wait, nevermind.
[22:55:18 CET] Action: mac_nibblet did not copy most of this of stackoverflow.
[23:00:35 CET] <mac_nibblet> Fenrirthviti, any idea how to do this ?
[23:00:46 CET] <Fenrirthviti> I do not.
[23:02:21 CET] <kcghost> Quick question regarding mapping audio streams. I am specifying by an id, e.g. "-map i:0x1100", and found that I got two audio streams in the output since there are two input audio streams with that id. Is there a way to tell ffmpeg to select only one? Perhaps the highest quality one?
[23:11:39 CET] <idlus> kcghost: you can select streams with "m:key[:value]" rather than "i:index", maybe you could use that to specify quality?
[23:16:32 CET] <kcghost> idlus: this is to avoid having to parse it out, I don't know the quality or much about the stream beforehand
[23:50:16 CET] <kcghost> idlus: I think I have a somewhat crazy solution, have ffmpeg with the ids and copy streams and pipe into another ffmpeg that will by default just select the best streams.
[23:51:11 CET] <kcghost> idlus: though i think I ran into a bug with mpegts and the 'truehd' audio codec, it shows up as an unknown stream in the output.
[00:00:00 CET] --- Thu Mar 23 2017