[Ffmpeg-devel-irc] ffmpeg.log.20170811

[burek]

[04:09:49 CEST] <djgl> Hi, I noticed that the yuv2rgb routines on x86 need (width+7)/8 <= stride/8. Otherwise they will drop pixels on the right. Is this documented somewhere?
[09:30:02 CEST] <Fyr> guys, can FFMPEG utilize NVIDIA Tesla?
[09:43:37 CEST] <squ> some encoders can use cuda
[09:44:26 CEST] <Fyr> squ, so h264_nvenc?
[09:44:47 CEST] <squ> yes
[09:46:10 CEST] <furq> nvenc doesn't use cuda
[09:46:18 CEST] <furq> some filters can use opencl
[09:46:31 CEST] <furq> and nvdec will use cuda if your card doesn't have the asic for it
[09:47:47 CEST] <Fyr> furq, is there a list of the filters that support openCL?
[09:47:55 CEST] <Fyr> I can't google them out.
[09:48:11 CEST] <Fyr> >>OpenCL is currently only used for filtering (deshake and unsharp filters)
[09:48:21 CEST] <Fyr> hwaccelintro says.
[09:48:36 CEST] <Fyr> I'm sure there is more.
[10:53:42 CEST] <furq> those are the only ones with opencl parameters
[10:54:10 CEST] <furq> there are definitely more filters which could potentially use it if you want to contribute patches
[10:54:31 CEST] <furq> i know there are some good opencl denoisers out there
[14:47:47 CEST] <under9000> hi guys, I have a problem here. normally I don't resort to asking people but 30 minutes of googling, docs searching and stackexchange browsing resulted in nothing. the question is: is there an ffmpeg command line argument that tells it to REMOVE output file if it failed to convert from one container to another?
[14:48:46 CEST] <furq> || rm out.mkv
[14:51:39 CEST] <under9000> would it work on windows?
[14:51:56 CEST] <DHE> in a batch file or powershell, you could make it work
[14:56:49 CEST] <under9000> ffmpeg is being launched from youtube-dl which tells it the output filename, and I don't know of a way to tell the output filename to rm command
[14:57:29 CEST] <under9000> since I can only pass additional arguments to youtube-dl which passes them on to ffmpeg
[14:58:05 CEST] <under9000> am I correct to assume that ffmpeg DOESN'T have a specific command switch that tells it to remove output file on error?
[15:13:49 CEST] <DHE> I can't find one in ffmpeg-all.html
[15:19:02 CEST] <under9000> ok thank you
[15:23:38 CEST] <dynek> hello!
[15:24:45 CEST] <dynek> I'm running 3.3.3-1 and trying to use va-api with h.264 video as source
[15:25:26 CEST] <dynek> I end up with a video split in two full or pinkness, let me upload a screenshot
[15:26:47 CEST] <dynek> ffmpeg output: https://pastebin.com/9LmR2EHW
[15:27:44 CEST] <BtbN> vaapi is pretty much untested with anything but intel
[15:29:26 CEST] <dynek> screenshot: https://cloud.bs-network.net/index.php/s/zcDb3R9CC1r4xAd
[15:29:50 CEST] <dynek> wanted to try vdpau but I need x11 to be running and wasnt successful using LXC
[15:39:24 CEST] <jkqxz> dynek:  That's made by the encoder on AMD?  Read the AMD section of <https://trac.ffmpeg.org/wiki/Hardware/VAAPI>, particularly the last paragraph.
[15:39:55 CEST] <dynek> jkqxz: it is - thank you I'll have a look
[15:41:45 CEST] <dynek> jkqxz: I owe you a beer :-)
[15:59:31 CEST] <dynek> is it possible to use va-api on the decoding side only (to avoid using cpu cycles) and only extract jpegs every seconds?
[15:59:45 CEST] <dynek> trying to do so but it's as consuming with or without hwaccel flags
[16:09:50 CEST] <jkqxz> Should be.  What is your command line at the moment?
[16:21:57 CEST] <truexfan81> so why does the ffmpeg configure script take so long?
[16:22:10 CEST] <c_14> because it tests a bunch of things
[16:23:52 CEST] <truexfan81> i wonder if there is an argument i can give it, to make it show what its checking, cause with it giving no output it makes it look like its frozen
[16:24:07 CEST] <c_14> tail -f ffbuild/config.log
[16:25:32 CEST] <truexfan81> my cpu is pegged at the moment, cause make is running with -j6
[16:39:51 CEST] <Guest80657> Hi there, does anybody have an example of transmuxing X to fragmented MP4? (it'd be better if it was ffmpeg/lib example but I'll be happy with ffmpeg command line as well).
[16:40:30 CEST] <Guest80657> I found a remuxing example (using the lib) https://ffmpeg.org/doxygen/trunk/doc_2examples_2remuxing_8c-example.html but then I realize I also need to understand how to mux to fragmented mp4 not only mp4.
[16:42:55 CEST] <dreampeppers99> I found some samples in the Internet telling to use `-movflags empty_moov+default_base_moof+omit_tfhd_offset` in order to obtain such fmp4
[16:44:16 CEST] <dreampeppers99> I think I found the info I need to make it through ffmpeg command line https://ffmpeg.org/ffmpeg-formats.html#mov_002c-mp4_002c-ismv
[16:44:51 CEST] <JEEB> yup
[16:45:05 CEST] <dreampeppers99> anyway, if anybody still has a link to an example of remuxing to fragmented mp4 using the ffmpeg lib it'd be great!
[17:24:52 CEST] <Fenrirthviti> Hmm, does VBR for NVENC actually use CQP at all?
[17:40:04 CEST] <leemox> Which is the best deinterlace filter right now? In the ffmpeg documentation I see that it is 5 different - bwdif, kerndeint, nnedi, w3fdif and yadif. Which one I will get the best results?
[17:40:15 CEST] <leemox> https://ffmpeg.org/ffmpeg-filters.html
[17:42:37 CEST] <furq> you forgot mcdeint
[17:43:03 CEST] <furq> nnedi is probably the best but it's also too slow to be usable in ffmpeg
[17:43:19 CEST] <furq> maybe mcdeint is good but when i tried it it was even slower than nnedi and looked terrible
[17:43:28 CEST] <furq> so the answer is probably bwdif=1
[17:46:02 CEST] <leemox> Thank you very much. What about yadif? In various forums it was quite often recommended.
[17:47:03 CEST] <furq> bwdif is supposedly better than yadif
[17:47:06 CEST] <furq> it's worth trying both though
[17:47:14 CEST] <furq> they looked basically the same last time i tested them
[17:47:42 CEST] <furq> just make sure you use them both in mode 1
[17:49:38 CEST] <leemox> Why is it so important? The default settings are different (mode 0 in bwdif). I thought to trust them :-)
[17:50:41 CEST] <furq> mode 1 is double framerate
[17:50:44 CEST] <furq> it looks much better
[17:57:31 CEST] <pgorley> hi, how do i get the number of samples in an AVStream?
[17:57:52 CEST] <leemox> Have you tested w3fdif?
[17:58:50 CEST] <JEEB> w3fdif is one of the two that bwdif uses (together with yadif)
[17:58:52 CEST] <JEEB> IIRC
[18:01:41 CEST] <leemox> Yes, indeed. Before asking previous questions, I wondered if the combination of several different methods works better than one. But as the @furq wrote, bwdif is reportedly very good.
[18:01:46 CEST] <leemox> So propably yes.
[18:04:07 CEST] <furq> does anyone know how to actually use mcdeint properly
[18:04:14 CEST] <furq> the results were so bad when i tried it that i suspect i did it wrong
[18:04:42 CEST] <iive> you need yadif included before mcdeint
[18:04:46 CEST] <furq> yeah i did that
[18:05:00 CEST] <furq> it ran at like 2fps on pal dvd and looked absolutely awful
[18:05:09 CEST] <furq> and also spat out a ton of warnings
[18:05:22 CEST] <iive> hum... haven't tried it recently.
[18:06:06 CEST] <furq> i think i was just doing yadif=1,mcdeint=slow,tff
[18:06:34 CEST] <furq> it's gone from my bash history now
[18:28:09 CEST] <crot> Anyone have any experience trans-coding content from an hdhomerun
[18:30:14 CEST] <DHE> I do. the device I have is an OTA to ethernet converter
[18:33:15 CEST] <crot> Did you have any problems getting an acceptable quality overall the the streams look good but have intermittent pixelation
[18:34:40 CEST] <DHE> have you watched the raw streams? antenna quality and location matters quite a bit. use the tools to measure the signal quality
[18:35:14 CEST] <DHE> the other thing to watch for is that CPU isn't maxed out. this is a realtime stream. if your quality settings are too high you may be dropping packets
[18:35:19 CEST] <crot> Yeah raw is good. signal levels all look great.
[18:36:28 CEST] <crot> udp://235.71.0.4:1234?overrun_nonfatal=1 -vcodec libx264 -tune zerolatency -preset veryfast -profile:v high -x264opts nal-hrd=cbr:force-cfr=1 -b:v 2500k -minrate 1500k -maxrate 3500k -bufsize 3500k -a53cc 1 -acodec ac3 -ab 448000 -af 'volume=5' -vbsf h264_mp4toannexb -f mpegts 'udp://239.91.0.4:1234?
[18:37:49 CEST] <DHE> I disapprove of zerolatency mode here. I know it adds another ~1-2 seconds of playback delay but the effect is has on quality is very substantial.
[18:38:27 CEST] <DHE> also there's a chance the audio you're receiving is already AC3. do you need the volume boost here?
[18:39:05 CEST] <crot> yeah for what ever reason the audio is coming across lower than other channels on average.
[18:39:36 CEST] <crot> Stream #0:0[0x31]: Video: mpeg2video (Main) ([2][0][0][0] / 0x0002), yuv420p(tv, bottom first), 1280x720 [SAR 1:1 DAR 16:9], Closed Captions, 59.94 fps, 59.94 tbr, 90k tbn, 119.88 tbc     Stream #0:1[0x34](eng): Audio: ac3 (AC-3 / 0x332D4341), 48000 Hz, 5.1(side), fltp, 448 kb/s
[18:40:40 CEST] <crot> that is the ffprobe of the source stream. Would you recommend deinterlacing?
[18:43:40 CEST] <Plebber> When speciifying bitrates, does 512k means the same thing as 512K?
[18:44:02 CEST] <Plebber> as in is k kilobits/s and K kilobytes/s?
[18:44:13 CEST] <Plebber> I can't find documentation for these abbreviations
[18:45:39 CEST] <c_14> I think K is bytes and k is bits
[18:46:33 CEST] <Blubberbub_> isnt the difference that one is 1000 and the other is 1024?
[18:47:12 CEST] <Blubberbub_> (thats what wikipedia says - no clue how that is implemented in ffmpeg)
[18:48:46 CEST] <Plebber> But what about ffmpeg?
[18:49:46 CEST] <Plebber> If I said make a video with a 1 megabyte per second bitrate, what would you write down for -b:v ?
[18:50:00 CEST] <Plebber> Same for a half a megabyte
[18:51:44 CEST] <c_14> use 1MiB just to be safe
[18:53:11 CEST] <drv> for ffmpeg, -b:v is always in bits, and k == K
[18:54:10 CEST] <c_14>  * @param numstr a string representing a number, may contain one of the International System number postfixes, for example 'K', 'M', 'G'. If 'i' is appended after the postfix, powers of 2 are used instead of powers of 10. The 'B' postfix multiplies the value by 8, and can be appended after another postfix or used alone. This allows using for example 'KB', 'MiB', 'G' and 'B' as postfix.
[18:54:54 CEST] <drv> ah, you're right - didn't know about the B suffix
[18:56:02 CEST] <drv> the k/K/M/... multipliers are listed in libavutil/eval.c - si_prefixes array
[19:54:20 CEST] <alexpigment> I was working with a TV station today (albeit indirectly) and they accepted 720i MPEG-PS files
[19:54:28 CEST] <alexpigment> I had long thought that 720i was a myth
[19:54:38 CEST] <alexpigment> Am I wrong? Is this a real standard?
[20:09:35 CEST] <kepstin> north america? I don't think it's permitted in ATSC broadcast standards, so they're probably just upscaling it to 1080i
[20:12:35 CEST] <kerio> interlaced video is evil
[20:21:05 CEST] <leemox> Which mode is better to use for deinterlacing? In bwdif is 2 modes - send_frame and send_field. With the second option, the frame rate is doubled. This is great, but there is not any minuses on image quality?
[20:21:31 CEST] <leemox> *video quality
[20:21:41 CEST] <leemox> In second option
[20:29:03 CEST] <kepstin> leemox: the mode that doesn't double the framerate is basically equivalent to using the mode which does, and then dropping half the frames.
[20:29:34 CEST] <kepstin> I think it might be faster, but that's it. no improvement to quality.
[20:31:12 CEST] <leemox> So if I understand correctly, it is definitely to take a chance with doubling the number of frames. Thank you :-)
[20:31:39 CEST] <kepstin> keep in mind that 30frames per second interlaced video is actually 60 fields per second
[20:31:53 CEST] <kepstin> so if you deinterlace and output 30 frames per second, you're losing temporal information
[20:33:49 CEST] <leemox> Is there another, better than bwdif deinterlacer in your opinion?
[20:33:59 CEST] <leemox> in ffmpeg?
[20:34:30 CEST] <kepstin> bwdif is probably the best in ffmpeg, excluding nnedi
[20:35:15 CEST] <kepstin> if you want to try nnedi, it might be better to use a different implementation (not ffmpeg) that e.g. supports multithreading, asm optimizations, or even gpu compute.
[20:36:55 CEST] <leemox> Nnedi is much better? I know that based on neural networks, but probably not everything that uses something with AI is the best ;-)
[20:38:44 CEST] <kepstin> nnedi is an interpolator only, it just guesses what the missing lines are from the current field only. It does do a really good job of this, though. But in some cases, bwdif will give better results.
[20:39:29 CEST] <kepstin> state of the art is probably in some of the avisynth scripts/plugins people have done which combine nnedi with motion compensation :/
[20:41:08 CEST] <kepstin> (like furq I've tried mcdeint, but haven't actually seen it do anything useful, and I dunno if I'm doing it right)
[20:42:27 CEST] <kepstin> I tend to deal with stuff that's telecined more often than interlaced, tho, so I haven't played around with deinterlacing all that much.
[20:43:28 CEST] <leemox> What about nlmeans?
[20:44:30 CEST] <leemox> Oups, this is denoise :-)
[20:46:39 CEST] <alexpigment> i always use the yadif deinterlacer for what it's worth
[20:47:14 CEST] <alexpigment> i haven't done a whole lot of scientific comparisons between bwdif, yadif, or nnedi, but I know that yadif is nowhere near as good as whatever DXVA2 on Windows uses to deinterlace during playback
[20:47:34 CEST] <alexpigment> I deal with mostly true interlaced content fwiw, rather than telecined material
[20:49:20 CEST] <BtbN> dxva2 uses the hardware deinterlacer of the card
[20:50:37 CEST] <alexpigment> BtbN, i thought that might the be case, but I wasn't entirely sure
[20:50:44 CEST] <alexpigment> I've always used Nvidia cards for the most part
[20:50:57 CEST] <alexpigment> but I've got one HTPC using Intel HD 4600
[20:51:01 CEST] <BtbN> nvidias motion adaptive deinterlacing is quite nice
[20:51:13 CEST] <alexpigment> yeah, you can tell they're doing more than just a simple bob
[20:51:23 CEST] <alexpigment> (as you'd expect them to)
[20:51:45 CEST] <alexpigment> but yeah, I don't think there's anything as nice in ffmpeg to my knowledge
[20:51:55 CEST] <alexpigment> I hear that qtgmc is good in avisynth, but I haven't yet tried it
[21:12:50 CEST] <harvhat> hello. I'm having some trouble using the amerge filter
[21:15:49 CEST] <harvhat> trying to mos 2 stereo sources down to 1 steteo source (ie. -filter_complex "[0][1] amerge" ) but the second source always comes through as mono. I can swap the sources but its always the second source
[21:16:25 CEST] <harvhat> *mix 2 stereo sources down to 1 stereo source.. that is
[21:22:29 CEST] <kepstin> harvhat: that's not what the amerge filter is for, you probably want to use amix instead.
[21:31:24 CEST] <Ingvix> Hey, I've been trying to convert mp4 to mkv while keeping all of the multiple subtitles and audio tracks intact and also adding an additional external sub to the package. It seems that I'm always doing something wrong so could someone just end my suddering and provide an example command where I could learn from?
[21:38:33 CEST] <harvhat> kepstin: Okay yes that worked, thanks! I was basing my attempt off the second example under "2 × stereo ’ stereo" on this page in the wiki: https://trac.ffmpeg.org/wiki/AudioChannelManipulation. (the first example works but seems convoluted). "amix" worked perfect a though :)
[21:40:04 CEST] <kepstin> Ingvix: can you pastebin the output of "ffmpeg -i file.mp4" on your input? And what format is the external sub file?
[21:42:51 CEST] <Ingvix> kepstin, https://pastebin.com/ZMvRKMj7 and the subtitle is in ass format
[21:44:21 CEST] <kepstin> Ingvix: alright, something like "ffmpeg -i kimi.mp4 -i kimi.ass -map 0 -map 1 -c copy kimi.mkv" will probably do it, then.
[21:45:12 CEST] <kepstin> Ingvix: the "-map" options here say to copy all the streams from input files 0 and 1 respectively, and -c copy means to not attempt re-encoding streams (so the dvd subs will be copied as-is; without that it would probably give an error)
[21:49:49 CEST] <Ingvix> kepstin, https://pastebin.com/b9pYFEte I've been getting this error with my previous attempts too
[21:50:41 CEST] <kepstin> oh, huh, ffmpeg doesn't want to copy those mp4 'dvd_subtitle' tracks into the mkv eh.
[21:52:19 CEST] <Ingvix> I think it'd be okay to convert them if that could make it work
[21:52:24 CEST] <Ingvix> to ass of what ever
[21:52:27 CEST] <Ingvix> or*
[21:52:48 CEST] <kepstin> you could try adding "-c:s:0 dvd_subtitle -c:s:1 dvd_subtitle -c:s:2 dvd_subtitle -c:s:3 dvd_subtitle" to get ffmpeg to run those through its encoder chain, might fix it
[21:52:59 CEST] <kepstin> dvd subtitles are images, you can't just "convert" them to a text format :)
[21:53:21 CEST] <Ingvix> oh, okay, I try that then
[21:56:18 CEST] <Ingvix> does it matter where in the command I add that string? Should it be before -c copy?
[21:56:47 CEST] <kepstin> I'd put it after that, but I don't think it would make a difference
[21:56:58 CEST] <kepstin> has to be after all the input files and before the output file
[21:57:19 CEST] <Ingvix> I see
[21:58:04 CEST] <Ingvix> Well now it's processing it. Thanks for your help and explanation
[22:05:39 CEST] <Ingvix> Though can I somehow define the language of the new sub? it didn't seem to take it from the file name
[22:08:30 CEST] <Ingvix> I found that in documentation. Let's see if I can succesfully use it
[22:09:05 CEST] <cbsrobot> Ingvix: should be something like -metadata:s:0 language=eng
[22:09:23 CEST] <Ingvix> yes, that's what I found out
[22:13:10 CEST] <Dan__> Hi there, I'm having some problems converting DNXHD QT into PNGs. Does anyone here have any experience with this?
[22:15:02 CEST] <cbsrobot> Dan__: be more specific
[22:24:12 CEST] <Dan__> ffmpeg -i "/Users/Moody/Desktop/PLATES/movieclip.mov" -y -r 24 -vf setsar=1:1 -start_number 1001 "/Users/Moody/Desktop/PLATES/output.%04d.png"
[22:25:14 CEST] <Dan__> I'm trying to convert a DNXHD QT into a PNG sequence - the PNGs are coming out looking more crushed than the QT - does anyone know what codec or present I could use to stop this happening, and export a truer PNG representation of the original source file?
[22:33:37 CEST] <kepstin> Dan__: please pastebin the *output* of that ffmpeg command
[22:33:51 CEST] <kepstin> it has info that we need to know in order to help you (stuff like video codecs, etc)
[22:34:07 CEST] <Dan__> Sorry kepstin, how do I do that?
[22:34:42 CEST] <kepstin> run the command, copy/paste the text from the terminal onto a pastebin site
[22:39:09 CEST] <cbsrobot> Dan__: try: ffmpeg -i "/Users/Moody/Desktop/PLATES/movieclip.mov" -y -start_number 1001 "/Users/Moody/Desktop/PLATES/output.%04d.png"
[22:39:27 CEST] <cbsrobot> no need to set sar
[22:40:04 CEST] <Dan__> https://pastebin.com/xRpKQq86
[22:41:59 CEST] <Dan__> That didn't seem to work cbsrobot, I got the following error: av_interleaved_write_frame(): Input/output error
[22:46:09 CEST] <Mateon1> Hi, is it possible to output a wav file without metadata? I need to pass it into another program, but it has hardcoded offsets and bails out on metadata chunks
[22:46:32 CEST] <Mateon1> AKA, the file should only contain the "fmt " and "data" chunks
[22:50:05 CEST] <drv> Mateon1: '-fflags +bitexact' should turn off most of the extra metadata chunks, if I remember correctly
[23:03:57 CEST] <kepstin> Dan__: are you sure the problem isn't just that your image viewing application doesn't handle 16bit per channel png correctly? :/
[23:05:43 CEST] <Dan__> It's possible, but a colleague who I am working with exported the same QT in PNG format, and the blacks look less lifted than in the version I've managed to get from ffmpeg
[23:05:43 CEST] <kepstin> you could try adding -pix_fmt rgb24 so it outputs 8bit per channel png instead, just for comparison
[23:06:21 CEST] <Dan__> I had -pix_fmt rgb24 in the code before, but it still generates the same looking QT
[23:06:32 CEST] <Dan__> *sorry, the same looking PNG
[23:07:03 CEST] <kepstin> note that according to the ffmpeg output, it claims that the input is "tv" (aka limited) range, that could cause issues if it's incorrect.
[23:08:24 CEST] <kepstin> if that's wrong, you would see crushed blacks and whites due to clipping on the conversion to rgb.
[23:09:38 CEST] <alexpigment> oh man, i've dealt with color/gamut/etc issues in the past. it's a nightmare. sometimes the video card settings are at play, sometimes the application is at play, sometimes (often) it's because the monitor is a wide gamut monitor with a ProPhoto or Adobe 98 color profile. good luck on this one guys :)
[23:10:16 CEST] <alexpigment> and yes, full 0-255 vs limited 16-235 is confusing
[23:10:27 CEST] <Dan__> I see - is there a way to stop the input being limited?
[23:10:52 CEST] <kepstin> this is 10bit, so it's even more confusing (i dunno what the exact ranges are there)
[23:10:57 CEST] <alexpigment> yep
[23:11:08 CEST] <alexpigment> especially considering most monitors don't actually support 10-bit
[23:11:16 CEST] <alexpigment> (or anything above it, obviously)
[23:11:26 CEST] <kepstin> Dan__: the issue is that metadata on the input is wrong. I think this can be overridden in ffmpeg, but I forget exactly how, might have to look it up.
[23:11:38 CEST] <alexpigment> -color_range full ?
[23:11:44 CEST] <alexpigment> i think i remember using that in the past
[23:12:10 CEST] <alexpigment> actually, i'm going to do as i said and stay out of this one
[23:12:23 CEST] <kepstin> -color_range pc (or -color_range jpeg) as an input option yeah
[23:12:42 CEST] <kepstin> have to try and see if it's actually accepted, it says it's just a "hint to the decoder"
[23:14:04 CEST] <kepstin> if adding that fixes the issue, then the problem is that whatever made your qt file to start with did it incorrectly :/
[23:20:02 CEST] <Dan__> hmm, it doesn't appear to be working, but I'm still tinkering
[23:20:20 CEST] <Dan__> pc or jpeg weren't being accepted, so I had a quick google and think it needs to be either -color_range 1, or -colour_range 2
[23:20:36 CEST] <Dan__> > The source says:  >  > 0 == unspecified  > 1 == "MPEG" range  > 2 == "JPEG" range  >  > Therefore, I'd expect 1 to be "broadcast" 16-235, and 2 to be "full"  > 0-255. I haven't tested this, just grepped through the source.
[23:20:49 CEST] <kepstin> it should accept those values
[23:21:05 CEST] <kepstin> you've only got ffmpeg 2.8 tho, you might want to grab a newer version...
[23:21:37 CEST] <Dan__> yes I will do, thanks for your help kepstin
[23:21:42 CEST] <kepstin> and what do you mean by "weren't being accepted" - did it actually give an error saying so?
[23:21:57 CEST] <Dan__> [png @ 0x7fe2d6802a00] [Eval @ 0x7fff59df6470] Undefined constant or missing '(' in 'pc' [png @ 0x7fe2d6802a00] Unable to parse option value "pc" [png @ 0x7fe2d6802a00] Error setting option color_range to value pc.
[23:22:34 CEST] <kepstin> that doesn't make sense...
[23:22:51 CEST] <kepstin> are you sure you used it as an *input* option (before the -i)?
[23:24:43 CEST] <Dan__> Sorry Kepstin you're absolutely right
[23:24:46 CEST] <Dan__> I've just done that and it's worked!
[23:24:50 CEST] <Dan__> Thank you so much.
[23:25:27 CEST] <Dan__> This will save me hours
[23:26:16 CEST] <kepstin> but yeah, this option is just a hack, it says "ignore what the input file claims to be, and treat it as if it was this instead" - it shouldn't be needed on most file.
[23:26:47 CEST] <viric> kepstin: I'm aligning things fine and correcting drift. with octave + fftconv + ffmpeg.
[23:26:50 CEST] <iive> Dan__: why do you want them as png?
[23:30:50 CEST] <Dan__> It's an export from the Avid and we need to work on image sequences in the software we use
[23:35:41 CEST] <Dan__> Well as a hack it does exactly what I need - thanks kepstin!
[23:36:07 CEST] <Ingvix> Can I convert ass to some image format with ffmpeg?
[23:36:27 CEST] <c_14> You con burn it onto a possibly transparent background
[23:36:44 CEST] <c_14> so, yeah I guess
[23:37:08 CEST] <c_14> You'll lose all timestamp info though (with image output)
[23:37:24 CEST] <c_14> use the ass filter (or the subtitles filter) onto a color source with 100% alpha or so
[23:38:47 CEST] <Ingvix> it's just that my smart tv does not like to use fonts, colors or positions determined in the ass file so I thought that it they could work if it was in some image based sub format
[23:40:20 CEST] <alexpigment> does anyone know why the -standard parameter doesn't work (or what its replacement is)?
[23:40:32 CEST] <alexpigment> if i use -tvstd ntsc , it tells me to use -standard instead
[23:40:42 CEST] <c_14> Ingvix: you can just burn it onto the video
[23:40:56 CEST] <c_14> I'm not entirely sure ffmpeg can convert text subs to video subs though
[23:41:03 CEST] <Ingvix> c_14, yes but I don't want to lose the quality
[23:41:14 CEST] <Ingvix> of the video
[23:41:23 CEST] <c_14> an x264 encode at decent bitrates should be (visually) lossless
[23:41:27 CEST] <c_14> unless your source is terrible
[23:41:54 CEST] <Ingvix> my source is 16gb hevc 4k
[23:42:18 CEST] <Ingvix> uh, the 16gb is just size of the video file but anyway
[23:43:12 CEST] <alexpigment> not that it matters really, but content is key here. 16gb for 1 minute of 4k = great quality, 16gb for 2 hours 4k = ok quality
[23:43:21 CEST] <alexpigment> er, *context* is key
[23:44:09 CEST] <alexpigment> still, c_14's point about being able to encode at something visually lossless is certainly possible, as long as you don't mind large file sizes
[23:44:46 CEST] <Ingvix> well it's anime and it's hevc so if I've understood it correctly it got almost twice the bitrate to h264
[23:44:49 CEST] <alexpigment> that said, encoding from hevc to hevc sounds like a bitch
[23:44:52 CEST] <Ingvix> or something like that
[23:45:00 CEST] <Ingvix> yes, I've done it once
[23:45:10 CEST] <Ingvix> and I have those other subs also that I would like to preserve, which are already dvd subtitles
[23:46:20 CEST] <c_14> I'm just going to throw this out there,
[23:46:28 CEST] <c_14> but just connect an HDMI cable from your computer to the TV?
[23:46:57 CEST] <alexpigment> i was just thinking the same thing. or invest in an HTPC that supports Kodi. worth it in the long run imho
[23:47:00 CEST] <Ingvix> well, the thing here is that my laptop does not offer 4k over hdmi
[23:47:27 CEST] <alexpigment> ingvix does it offer 4k over displayport?
[23:47:39 CEST] <Ingvix> it does not have displayport
[23:47:43 CEST] <alexpigment> k
[23:48:27 CEST] <Ingvix> so creating a good vobsub or what ever out of ass it not possible?
[23:48:37 CEST] <c_14> It's technically possible
[23:48:58 CEST] <c_14> You could maybe give SubtitleEdit a try
[23:49:08 CEST] <Ingvix> I shall
[23:49:27 CEST] <c_14> >Export to PNG images (+bdn xml), Adobe Encore FAB image script, VobSub, Blu-ray sup, EBU stl, PAC, and plain text
[23:49:36 CEST] <c_14> so should be able to export to VobSub and PGS
[23:50:24 CEST] <Ingvix> and ffmpeg is able to attach them to the movie?
[23:50:30 CEST] <c_14> yeah
[23:50:34 CEST] <c_14> that's the easy part
[23:50:39 CEST] <Ingvix> yes, probably
[23:51:28 CEST] <Ingvix> also I tried to burn the subs to the file before on this journey to get things right but the output file was only something over 6gb
[23:51:48 CEST] <alexpigment> Ingvix: what did you set the crf or bitrate to?
[23:51:59 CEST] <alexpigment> and did the quality look noticeably different?
[23:52:18 CEST] <alexpigment> sometimes you reach saturation at a lower bitrate. that's especially true with animation
[23:52:32 CEST] <Ingvix> I don't think I set anything like that. I didn't even bother to check the quality...
[23:52:38 CEST] <Ingvix> but I guess I'm a bit smarter now then
[23:52:43 CEST] <c_14> crf 23 probably then
[23:53:19 CEST] <Ingvix> but I try the subtitleedit for starters
[23:56:31 CEST] <alexpigment> throwing this out there again:
[23:56:38 CEST] <alexpigment> does anyone know why -standard doesn't work in FFMPEG? if I use -tvstd ntsc, it tells me to use -standard. if i use -standard ntsc, it says standard is an unrecognized option
[00:00:00 CEST] --- Sat Aug 12 2017