[Ffmpeg-devel-irc] ffmpeg.log.20191230

[burek]

[00:37:09 CET] <squaresmile> I'm trying to play criware's usm video. I used the VGMToolbox to extract the audio and video streams. The audio is HCA (not ADX) and plays fine so I'm inclined to believe the extraction was correct.
[00:37:17 CET] <squaresmile> However, the audio seems to be h264 (not MPEG) and I can't get it to play or convert. Here's the ffprobe https://pastebin.com/raw/6Z9ydNSU and mediainfo https://pastebin.com/raw/RcHkFZdn
[00:37:45 CET] <squaresmile> *the video
[00:38:04 CET] <squaresmile> Any help to play the file would be appreciated
[00:39:58 CET] <nicolas17> h264 is a video codec
[00:40:38 CET] <nicolas17> oh I see your correction now
[00:40:55 CET] <nicolas17> my eyes are tired enough for the day already x_x
[01:12:39 CET] <cehoyos> Please provide a sample file
[01:18:05 CET] <squaresmile> https://www.mediafire.com/file/px4u7so4wqpx6zj/ops00100_40534656.m2v/file
[01:22:05 CET] <squaresmile> It should decode to something like this https://www.youtube.com/watch?v=7DvRo4o6rpw
[01:25:38 CET] <cehoyos> This file is not playable, the extraction did not work correctly
[01:40:55 CET] <squaresmile> hmm, I guess back to the extraction. Since the audio extracted alright, I thought it was just some messed up h264 proprietary encode
[01:42:11 CET] <cehoyos> This may also be possible
[03:04:26 CET] <tanlis> Hi all, I kicked off a download of a HLS stream(m3u8) using youtube-dl. It should of ended by now(football game), but the stream is still broadcasting and thus FFmpeg is still downloading data. What is the best way to force FFmpeg to stop download and begin converting/muxing the file into a viewable format? If I try Ctrl-C, it just results in a
[03:04:26 CET] <tanlis> broken mp4.part file that doesn't seem to be viewable by anything. Running 4.2.1 of FFmpeg. Appreciate the help.
[03:08:44 CET] <nicolas17> mp4 has an index that needs to be written at the end of the file
[03:09:14 CET] <nicolas17> I thought Ctrl-C would let ffmpeg cleanly shutdown and do that finalization
[03:09:41 CET] <nicolas17> but maybe not when it's running under youtube-dl? idk
[03:10:02 CET] <tanlis> That's what I keep reading, but when I try it, the file just ends up as the mp4.part and I see something like:
[03:10:03 CET] <tanlis> frame=  900 fps= 88 q=-1.0 Lsize=   16214kB time=00:00:30.04 bitrate=4420.5kbits/s speed=2.95x    video:16162kB audio:20kB subtitle:0kB other streams:0kB global headers:0kB muxing overhead: 0.196695%Exiting normally, received signal 2.
[03:10:31 CET] <tanlis> And everything on the youtube-dl GitHub site, says to do ctrl-c
[03:11:58 CET] <nicolas17> try running ffprobe on the mp4.part file
[03:14:41 CET] <tanlis> Hrm, so maybe this will work. I do see this in the test file when I ran ffprobe. Input #0, mov,mp4,m4a,3gp,3g2,mj2, from 'giants-giants.mp4.part':
[03:14:48 CET] <tanlis> And when I renamed the file to .mp4, that seemed to work
[03:15:14 CET] <nicolas17> oh nice, so ffmpeg did finalize the file, just youtube-dl didn't rename it
[03:15:37 CET] <nicolas17> does it play when renamed?
[03:15:54 CET] <tanlis> The test file did. Haven't tried it on the full file.
[03:16:35 CET] <tanlis> About to. Just doing a little more testing
[03:22:33 CET] <tanlis> That worked. Phew
[03:22:40 CET] <tanlis> Thanks nicolas17
[03:23:31 CET] <nicolas17> :)
[04:39:30 CET] <Hello71> I'm inclined to guess that it has empty moov or something
[04:43:45 CET] <nicolas17> Hello71: yeah, seems Ctrl-Cing made ffmpeg write the full moov at the end, but didn't make youtube-dl rename the file
[04:43:50 CET] <nicolas17> so it was still a .part
[04:50:42 CET] <Hello71> maybe
[04:50:55 CET] <Hello71> but it seems more likely that the file is saved as streamable by default
[04:51:06 CET] <Hello71> so you can do youtube-dl & mpv
[04:51:49 CET] <nicolas17> how do you save an .mp4 as streamable?
[04:51:57 CET] <nicolas17> fragments?
[05:26:05 CET] <phk> hi guys, are scale_cuda and scale_npp filters still supports? because it not working for me while I have nvenc/cuvid encoder decoders support
[09:52:57 CET] <OverCoder> hi
[09:53:12 CET] <OverCoder> I'm trying to capture screen using x11grab but with Intel QuickSync
[09:55:23 CET] <OverCoder> this totally works https://storage.overcoder.dev/irc/614e19c44ae0b3b4/image.png
[09:55:55 CET] <OverCoder> er, I meant this  https://storage.overcoder.dev/irc/aa053bccf55fcff4/image.png  does work
[09:56:13 CET] <OverCoder> but as soon as I switch h264 to h264_qsv, I get this:  https://storage.overcoder.dev/irc/3175f1b8be0c1964/image.png
[09:56:49 CET] <OverCoder> I tried going parameter by parameter migration a copy pasted ffmpeg encoding command from docs to work with x11grab - apparently I can't use QSV at all when I use x11grab? or am I missing something?
[09:56:52 CET] <OverCoder> For the matter, I use vaapi
[14:02:16 CET] <Etua> Hello, I'd like to write simple C program that would sort audio files using AcoustID/MusicBrainz database. AcoustID has Chromaprint - a C library which can tag already decoded audio streams, but I think that instead of externally decoding files I could wrap FFmpeg to decode files using demuxer and use Chromaprint muxer to tag them with AcoustID ID. I am not really familiar with FFmpeg architecture so I am not sure whether I get it right.
[14:37:43 CET] <vlt> durandal_1707: Thank you! Using "fillborders=bottom=2" *before* "perspective=..." worked great :)
[14:41:46 CET] <vlt> How can I get a clip of a specific duration from a single image? I tried `ffmpeg -f image2 -i still.png -r 25 -t 5 ...` but get an output with "Duration: 00:00:00.04" (only one frame). What do I need to add?
[14:46:36 CET] <vlt> durandal_1707: (with "mode=fixed", of course, to get a black border)
[14:47:17 CET] <furq> vlt: -loop 1 -i still.png
[14:51:00 CET] <vlt> furq: Yes! Thans :)
[17:28:18 CET] <phk> hi guys, are scale_cuda and scale_npp filters still supports
[17:28:40 CET] <phk> i dont see any of those filters in -filters list
[17:34:18 CET] <Etua> Hello, I'd like to write simple C program that would sort audio files using AcoustID/MusicBrainz database. AcoustID has Chromaprint - a C library which can tag already decoded audio streams, but I think that instead of externally decoding files I could wrap FFmpeg to decode files using demuxer and use Chromaprint muxer to tag them with AcoustID ID. I am not really familiar with FFmpeg architecture so I am not sure whether I get it right.
[17:39:06 CET] <durandal_1707> phk: need to compile own version
[17:40:24 CET] <phk> durandal_1707: according to this https://trac.ffmpeg.org/wiki/HWAccelIntro
[17:40:39 CET] <phk> we just need to install nv-codec-headers
[17:40:43 CET] <phk> right?
[17:41:24 CET] <phk> or still need to enable anything in ffmpeg compilation time
[17:44:28 CET] <phk> and if something else i have to do thn can you please tell me more?
[17:45:40 CET] <pink_mist> you always need to eanble things in compulation
[17:48:27 CET] <phk> pink_mist: afaik, nvenc/nvdec are autodetected at ffmpeg configuration time
[17:49:29 CET] <phk> but im afraid, im wrong if scale_npp are from different package
[17:50:21 CET] <phk> because i see nvdec and nvenc in encoders/decoders list
[17:50:53 CET] <phk> but dont see scale_npp or cuda_scale
[17:51:48 CET] <BtbN> It's called scale_cuda
[17:52:08 CET] <BtbN> And you need clang or the CUDA sdk for cuda filters to be built.
[18:12:39 CET] <kepstin> Etua: such a tool is certainly possible, but you can't do it in a single pass with ffmpeg. Note also that chromaprint and acoustid are not the same thing - you need to send the chromaprint to a web service, and that responds with the acoustid to tag.
[18:13:57 CET] <kepstin> Etua: You can use ffmpeg to generate the chromaprint, but while you can use ffmpeg to do the second pass to add the metadata, it might make sense to use taglib or something instead, since it can (sometimes) add the metadata in-place.
[18:14:48 CET] <kepstin> Etua: also, feel free to join us in #metabrainz if you want to talk about musicbrainz tagging tools :)
[18:17:19 CET] <Etua> kepstin: Thank you, I'll ask the follow-up question there.
[19:33:02 CET] <phk> im getting "Impossible to convert between the formats supported by the filter 'Parsed_scale_cuda_0' and the filter 'auto_scaler_0'" error in filter parsing
[19:33:16 CET] <phk> in command: ffmpeg -hwaccel cuvid -c:v mpeg2_cuvid -i udp://239.255.1.10:1234 -vf scale_cuda=w=1920:h=1080 -f mpegts -y scale_cuda.ts -loglevel debug
[19:33:20 CET] <phk> log: https://pastebin.com/udC87Evr
[19:33:21 CET] <phk> filter_log: https://pastebin.com/t5dEEZUd
[19:47:05 CET] <kepstin> phk: you're using a software encoder, so you have to use the hwdownload filter to transfer the frame into cpu memory.
[19:47:18 CET] <kepstin> (or switch to using a hardware encoder, i suppose)
[19:52:12 CET] <phk> kepstin: thanks it working in both ways. hwupload/hwdownload and setting hwencoder
[19:54:22 CET] <phk> now problem fall in mpv :(
[19:54:27 CET] <phk> not working in mpv
[23:14:13 CET] <iconoclasthero> [ipod @ 0x564479a4e580] Frame rate very high for a muxer not efficiently supporting it.
[23:14:13 CET] <iconoclasthero> Please consider specifying a lower framerate, a different muxer or -vsync 2
[23:14:13 CET] <iconoclasthero> [libx264 @ 0x564479a4b140] height not divisible by 2 (160x137)
[23:14:13 CET] <iconoclasthero> Error initializing output stream 0:0 -- Error while opening encoder for output stream #0:0 - maybe incorrect parameters such as bit_rate, rate, width or height
[23:14:31 CET] <iconoclasthero> Command line:  "for i in *.flac; do ffmpeg -i "$i" -acodec libfdk_aac -vbr 4 "${i%.*}.m4a"; done"
[23:14:38 CET] <iconoclasthero> how do i fix this?
[23:17:56 CET] <iconoclasthero> https://pastebin.com/jPG253uz
[23:22:54 CET] <furq> iconoclasthero: -vn
[23:23:25 CET] <furq> or actually -c copy should work for m4a
[23:23:34 CET] <furq> -c:v copy rather
[23:29:23 CET] <iconoclasthero> there's no video in it.  ir's a flac audio...
[23:41:10 CET] <pink_mist> well why are you trying to use x264 for audio?!
[23:41:40 CET] <iconoclasthero> i'm not
[23:42:00 CET] <pink_mist> then the output you showed us here in the channel is probably a lie
[23:42:15 CET] <pink_mist> it clearly mentions libx264
[23:43:45 CET] <pink_mist> why don't you just try furq's suggestion?!
[23:48:40 CET] <iconoclasthero> i did, it worked...but i'm not 100% sure it will play on the car stereo.  can i drop the 0:0 stream some how?
[23:51:45 CET] <iconoclasthero> and also, why would ffmpeg lie in the output?
[23:51:57 CET] <iconoclasthero> it appeared to me that the video channel might have been an image?
[00:00:00 CET] --- Tue Dec 31 2019