October 2024 - ffmpeg-user - lists.ffmpeg.org

has setpts filter been changed?
by William C Bonner 21 Mar '25

21 Mar '25

I use setpts to speed up some video that I'm concatenating from my car's dashcam. The process is consistent and hasn't changed recently. Here's an example command line: C:\Windows\ffmpeg.exe -hide_banner -i D:\ROAV\MOVIE\2024_0803_173613_229.MP4 -i D:\ROAV\MOVIE\2024_0803_173713_230.MP4 -i D:\ROAV\MOVIE\2024_0803_173813_231.MP4 -i D:\ROAV\MOVIE\2024_0803_173913_232.MP4 -i D:\ROAV\MOVIE\2024_0803_174013_233.MP4 -i D:\ROAV\MOVIE\2024_0803_174113_234.MP4 -i D:\ROAV\MOVIE\2024_0803_174213_235.MP4 -i D:\ROAV\MOVIE\2024_0803_174313_236.MP4 -i D:\ROAV\MOVIE\2024_0803_174413_237.MP4 -i D:\ROAV\MOVIE\2024_0803_174513_238.MP4 -i D:\ROAV\MOVIE\2024_0803_174613_239.MP4 -i D:\ROAV\MOVIE\2024_0803_174713_240.MP4 -i D:\ROAV\MOVIE\2024_0803_174813_241.MP4 -i D:\ROAV\MOVIE\2024_0803_174913_242.MP4 -i D:\ROAV\MOVIE\2024_0803_175013_243.MP4 -i D:\ROAV\MOVIE\2024_0803_175113_244.MP4 -i D:\ROAV\MOVIE\2024_0803_175213_245.MP4 -i D:\ROAV\MOVIE\2024_0803_175313_246.MP4 -i D:\ROAV\MOVIE\2024_0803_175413_247.MP4 -i D:\ROAV\MOVIE\2024_0803_175513_248.MP4 -i D:\ROAV\MOVIE\2024_0803_175613_249.MP4 -i D:\ROAV\MOVIE\2024_0803_175713_250.MP4 -i D:\ROAV\MOVIE\2024_0803_175813_251.MP4 -i D:\ROAV\MOVIE\2024_0803_175913_252.MP4 -i D:\ROAV\MOVIE\2024_0803_180013_253.MP4 -i D:\ROAV\MOVIE\2024_0803_180113_254.MP4 -i D:\ROAV\MOVIE\2024_0803_180213_255.MP4 -filter_complex [0:v][1:v][2:v][3:v][4:v][5:v][6:v][7:v][8:v][9:v][10:v][11:v][12:v][13:v][14:v][15:v][16:v][17:v][18:v][19:v][20:v][21:v][22:v][23:v][24:v][25:v][26:v]concat=n=27:v=1[v];[v]setpts=(1/60)*PTS,drawtext=fontfile=C\\:/Windows/Fonts/consola.ttf:fontcolor=white:fontsize=80:y=main_h-text_h-50:x=50:text=WimsWorld[o] -map [o] -map_metadata -1 -c:v libx265 -crf 23 -preset veryfast -movflags +faststart -bf 2 -g 15 -pix_fmt yuv420p -n D:\ROAV\20240803-BMW.mp4 The important part is setpts=(1/60)*PTS which should convert 1 minute of the source video to 1 second in the destination I noticed that watching recent videos strangely seemed much smoother, then realized that they were also significantly larger. https://trac.ffmpeg.org/wiki/How%20to%20speed%20up%20/%20slow%20down%20a%20… still says that using this method will drop frames, but when I ran ffprobe on the current file I got: C:\Users\Wim>ffprobe d:\ROAV\20240803-BMW.mp4 ffprobe version 2024-07-18-git-fa5a605542-full_build-www.gyan.dev Copyright (c) 2007-2024 the FFmpeg developers built with gcc 13.2.0 (Rev5, Built by MSYS2 project) configuration: --enable-gpl --enable-version3 --enable-static --disable-w32threads --disable-autodetect --enable-fontconfig --enable-iconv --enable-gnutls --enable-libxml2 --enable-gmp --enable-bzlib --enable-lzma --enable-libsnappy --enable-zlib --enable-librist --enable-libsrt --enable-libssh --enable-libzmq --enable-avisynth --enable-libbluray --enable-libcaca --enable-sdl2 --enable-libaribb24 --enable-libaribcaption --enable-libdav1d --enable-libdavs2 --enable-libopenjpeg --enable-libquirc --enable-libuavs3d --enable-libxevd --enable-libzvbi --enable-libqrencode --enable-librav1e --enable-libsvtav1 --enable-libvvenc --enable-libwebp --enable-libx264 --enable-libx265 --enable-libxavs2 --enable-libxeve --enable-libxvid --enable-libaom --enable-libjxl --enable-libvpx --enable-mediafoundation --enable-libass --enable-frei0r --enable-libfreetype --enable-libfribidi --enable-libharfbuzz --enable-liblensfun --enable-libvidstab --enable-libvmaf --enable-libzimg --enable-amf --enable-cuda-llvm --enable-cuvid --enable-dxva2 --enable-d3d11va --enable-d3d12va --enable-ffnvcodec --enable-libvpl --enable-nvdec --enable-nvenc --enable-vaapi --enable-libshaderc --enable-vulkan --enable-libplacebo --enable-opencl --enable-libcdio --enable-libgme --enable-libmodplug --enable-libopenmpt --enable-libopencore-amrwb --enable-libmp3lame --enable-libshine --enable-libtheora --enable-libtwolame --enable-libvo-amrwbenc --enable-libcodec2 --enable-libilbc --enable-libgsm --enable-libopencore-amrnb --enable-libopus --enable-libspeex --enable-libvorbis --enable-ladspa --enable-libbs2b --enable-libflite --enable-libmysofa --enable-librubberband --enable-libsoxr --enable-chromaprint libavutil 59. 28.100 / 59. 28.100 libavcodec 61. 10.100 / 61. 10.100 libavformat 61. 5.101 / 61. 5.101 libavdevice 61. 2.100 / 61. 2.100 libavfilter 10. 2.102 / 10. 2.102 libswscale 8. 2.100 / 8. 2.100 libswresample 5. 2.100 / 5. 2.100 libpostproc 58. 2.100 / 58. 2.100 Input #0, mov,mp4,m4a,3gp,3g2,mj2, from 'd:\ROAV\20240803-BMW.mp4': Metadata: major_brand : isom minor_version : 512 compatible_brands: isomiso2mp41 encoder : Lavf61.5.101 Duration: 00:00:27.07, start: 0.000000, bitrate: 889896 kb/s Stream #0:0[0x1](und): Video: hevc (Main) (hev1 / 0x31766568), yuv420p(tv, progressive), 2560x1440, 889591 kb/s, 1800.04 fps, 1000k tbr, 1000k tbn (default) Metadata: handler_name : VideoHandler vendor_id : [0][0][0][0] encoder : Lavc61.10.100 libx265 Running it on an older file produced: C:\Users\Wim>ffprobe d:\ROAV\20240523-BMW.mp4 ffprobe version 2024-07-18-git-fa5a605542-full_build-www.gyan.dev Copyright (c) 2007-2024 the FFmpeg developers built with gcc 13.2.0 (Rev5, Built by MSYS2 project) configuration: --enable-gpl --enable-version3 --enable-static --disable-w32threads --disable-autodetect --enable-fontconfig --enable-iconv --enable-gnutls --enable-libxml2 --enable-gmp --enable-bzlib --enable-lzma --enable-libsnappy --enable-zlib --enable-librist --enable-libsrt --enable-libssh --enable-libzmq --enable-avisynth --enable-libbluray --enable-libcaca --enable-sdl2 --enable-libaribb24 --enable-libaribcaption --enable-libdav1d --enable-libdavs2 --enable-libopenjpeg --enable-libquirc --enable-libuavs3d --enable-libxevd --enable-libzvbi --enable-libqrencode --enable-librav1e --enable-libsvtav1 --enable-libvvenc --enable-libwebp --enable-libx264 --enable-libx265 --enable-libxavs2 --enable-libxeve --enable-libxvid --enable-libaom --enable-libjxl --enable-libvpx --enable-mediafoundation --enable-libass --enable-frei0r --enable-libfreetype --enable-libfribidi --enable-libharfbuzz --enable-liblensfun --enable-libvidstab --enable-libvmaf --enable-libzimg --enable-amf --enable-cuda-llvm --enable-cuvid --enable-dxva2 --enable-d3d11va --enable-d3d12va --enable-ffnvcodec --enable-libvpl --enable-nvdec --enable-nvenc --enable-vaapi --enable-libshaderc --enable-vulkan --enable-libplacebo --enable-opencl --enable-libcdio --enable-libgme --enable-libmodplug --enable-libopenmpt --enable-libopencore-amrwb --enable-libmp3lame --enable-libshine --enable-libtheora --enable-libtwolame --enable-libvo-amrwbenc --enable-libcodec2 --enable-libilbc --enable-libgsm --enable-libopencore-amrnb --enable-libopus --enable-libspeex --enable-libvorbis --enable-ladspa --enable-libbs2b --enable-libflite --enable-libmysofa --enable-librubberband --enable-libsoxr --enable-chromaprint libavutil 59. 28.100 / 59. 28.100 libavcodec 61. 10.100 / 61. 10.100 libavformat 61. 5.101 / 61. 5.101 libavdevice 61. 2.100 / 61. 2.100 libavfilter 10. 2.102 / 10. 2.102 libswscale 8. 2.100 / 8. 2.100 libswresample 5. 2.100 / 5. 2.100 libpostproc 58. 2.100 / 58. 2.100 Input #0, mov,mp4,m4a,3gp,3g2,mj2, from 'd:\ROAV\20240523-BMW.mp4': Metadata: major_brand : isom minor_version : 512 compatible_brands: isomiso2mp41 encoder : Lavf61.3.103 Duration: 00:09:40.88, start: 0.000000, bitrate: 26675 kb/s Stream #0:0[0x1](und): Video: hevc (Main) (hev1 / 0x31766568), yuv420p(tv, progressive), 2560x1440, 26672 kb/s, 25 fps, 25 tbr, 12800 tbn (default) Metadata: handler_name : VideoHandler vendor_id : [0][0][0][0] encoder : Lavc61.5.104 libx265 The important difference is that the older file reports 2560x1440, 26672 kb/s, 25 fps, 25 tbr, 12800 tbn and the newer file reports 2560x1440, 889591 kb/s, 1800.04 fps, 1000k tbr, 1000k tbn. This seems that the new code speed up the reported fps without dropping any frames.

3 4

Cannot seek an HLS stream with epoch timecode
by George Daramouskas 17 Nov '24

17 Nov '24

Hi everyone, I have an srt stream running 24/7, at some point in time I'd like to start an HLS stream that has as its input the srt stream. Now, the HLS stream could go down for arbitrary reasons and I am fine with that, it should be able to be restarted and keep going, however because I will be logging some events that take place at a point in time, I need to be able to refer to that point in time regardless of when the HLS stream started. This has led me to using unix epoch as a starting offset, specifically: ` ffmpeg -re -i srt://$url/?latency=20&mode=caller <http://10.21.6.126:9001/?latency=20&mode=caller> -map 0:v -map 0:a -hls_list_size 0 -master_pl_name stream1.m3u8 -strftime 1 -hls_segment_filename /core/data/stream1/%Y%m%d/segment-%s.ts -strftime_mkdir 1 -output_ts_offset $(date +%s) -hls_flags program_date_time /core/data/stream1.m3u8 ` Using ffprobe, I am able to verify that pts time is correct and gets incremented correctly: ` ffprobe -show_packets $url ........ [PACKET] codec_type=audio stream_index=1 pts=83020451563104 pts_time=1729592740.898000 dts=83020451563104 dts_time=1729592740.898000 duration=N/A duration_time=N/A size=344 pos=9420493 flags=K_ [/PACKET] ..... ` pts_time and dts_time look correct, same for the rest of the a/v streams. Now I want to seek on that hls and cut chunks out of it: ` ./ffmpeg_g -y -ss 1729592740.898000 -t 2 -seek_timestamp 1 -i https://10.21.6.110/stream1.m3u8 /tmp/test123.mp4 ` I am getting `could not seek to position 1729592740.898` and I end up with an improper mp4 cut from live instead from the right timestamp. Am I misusing the cli somehow here? Looking forward to hearing your thoughts Kind regards, George Daramouskas

2 1

Problem rewrapping DV in MOV
by Ganesh.Irelan＠syda.org 05 Nov '24

05 Nov '24

Hello I have .dv files captured from DV videotapes with vrecord. The .dv files play fine in Quicktime Player 10 but are not supported by Premier Pro or FCP. I want to wrap the DV stream from the .dv files into Quicktime MOV (without transcoding) so they can be imported into Premier Pro. FFMPEG rewraps the stream but when played in Quicktime, VLC, and Premier Pro there are narrow black vertical bars on the left and right edges of the frame. It looks like the production aperture 720x480 is being displayed instead of the clean aperture 704x480 but that’s just a guess. Can someone tell me an FFMPEG command that will result in MOV wrapped DV that displays without the vertical borders? This is what I tried. 3137-Mac-Pro-DAM:~ kirelan$ ffmpeg -i 36058.dv -map 0 -c copy 36058_Rewrap.mov ffmpeg version 6.1.1-tessus https://evermeet.cx/ffmpeg/ Copyright (c) 2000-2023 the FFmpeg developers built with Apple clang version 11.0.0 (clang-1100.0.33.17) configuration: --cc=/usr/bin/clang --prefix=/opt/ffmpeg --extra-version=tessus --enable-avisynth --enable-fontconfig --enable-gpl --enable-libaom --enable-libass --enable-libbluray --enable-libdav1d --enable-libfreetype --enable-libgsm --enable-libmodplug --enable-libmp3lame --enable-libmysofa --enable-libopencore-amrnb --enable-libopencore-amrwb --enable-libopenh264 --enable-libopenjpeg --enable-libopus --enable-librubberband --enable-libshine --enable-libsnappy --enable-libsoxr --enable-libspeex --enable-libtheora --enable-libtwolame --enable-libvidstab --enable-libvmaf --enable-libvo-amrwbenc --enable-libvorbis --enable-libvpx --enable-libwebp --enable-libx264 --enable-libx265 --enable-libxavs --enable-libxml2 --enable-libxvid --enable-libzimg --enable-libzmq --enable-libzvbi --enable-version3 --pkg-config-flags=--static --disable-ffplay libavutil 58. 29.100 / 58. 29.100 libavcodec 60. 31.102 / 60. 31.102 libavformat 60. 16.100 / 60. 16.100 libavdevice 60. 3.100 / 60. 3.100 libavfilter 9. 12.100 / 9. 12.100 libswscale 7. 5.100 / 7. 5.100 libswresample 4. 12.100 / 4. 12.100 libpostproc 57. 3.100 / 57. 3.100 [dv @ 0x7fbee8d06b80] Detected timecode is invalid [dv @ 0x7fbee8d06b80] Estimating duration from bitrate, this may be inaccurate Input #0, dv, from '36058_PRM.dv': Duration: 00:39:07.91, start: 0.000000, bitrate: 28771 kb/s Stream #0:0: Video: dvvideo, yuv411p, 720x480 [SAR 8:9 DAR 4:3], 28771 kb/s, 60k fps, 29.97 tbr, 60k tbn Stream #0:1: Audio: pcm_s16le, 48000 Hz, stereo, s16, 1536 kb/s Output #0, mov, to '36058_Rewrap_Map.mov': Metadata: encoder : Lavf60.16.100 Stream #0:0: Video: dvvideo (dvcp / 0x70637664), yuv411p, 720x480 [SAR 8:9 DAR 4:3], q=2-31, 28771 kb/s, 60k fps, 29.97 tbr, 60k tbn Stream #0:1: Audio: pcm_s16le (sowt / 0x74776F73), 48000 Hz, stereo, s16, 1536 kb/s Stream mapping: Stream #0:0 -> #0:0 (copy) Stream #0:1 -> #0:1 (copy) Press [q] to stop, [?] for help [out#0/mov @ 0x7fbee8d0b940] video:8246133kB audio:439852kB subtitle:0kB other streams:0kB global headers:0kB muxing overhead: 0.017190% size= 8687478kB time=00:39:07.87 bitrate=30311.5kbits/s speed=14.8x Thank you

3 5

Can't generate an .xml file
by Aditya Dandavate 02 Nov '24

02 Nov '24

ffmpeg -i distorted.mp4 -i original.mp4 -filter_complex libvmaf -f null - This command doesn't ouput any .xml file. How to tweak this command to get an .xml file ?

2 2

Can this trigger VBR encoding for x264, x265, and libvvenc ?
by Aditya Dandavate 01 Nov '24

01 Nov '24

If I set " -b:v 5000k -maxrate 6000k -minrate 4000k " for x264, x265 and vvenc, will this enable VBR encoding for these three encoders ?

2 9

Overlapping frames issue with -seek_frames_individually set to false
by Alex MacLean 30 Oct '24

30 Oct '24

Hello, I am struggling with ffmpeg to generate thumbnails for video files. We have to use `-seek_streams_individually false` because setting it to true takes far too long to retrieve frames near the end of longer content (> 30s runtime for content > 1h30m). Here is the command we currently have: ffmpeg -y -nostdin -copyts -seek_streams_individually false -ss 15.1 -i COMMCON.mp4 -vframes 1 -vsync 0 -c:v libwebp output.webp And the resulting thumbnail is attached (we have found overlapping frames around frames 451-456 for this content). What do you think? Any idea how we might be able to fix this issue and get a single frame in the result? Thank you, Alex MacLean

2 1

Intel Transcoding: Vaapi or QSV
by Shane Warren 30 Oct '24

30 Oct '24

I decided to start a new thread since I have done a lot of testing and patching on qsv and vaapi transcoding since last week. First, I decided to try to use QSV decoding and QSV encoding, but I noticed a53 captions were being stripped out, and I narrowed this down to the qsv decoders not passing the a53 captions on. I noticed there was a patch for SEI parsing for QSV decoders that was never merged into mainline ffmpeg. The patchset can be seen here: https://patchwork.ffmpeg.org/project/ffmpeg/list/?series=&submitter=&state=… I applied that patchset up to v6 of the patch. I had to fix a lot of API changes from that older version up to my mostly recent version but after that patch, QSV decode -> QSV encode passed a53 captions and they looked correct in my extensive testing (at least for h264). I'm trying to replace our use of nvidia cards and replace them with intel cards for transcoding. We make use of overlays, so I was trying to use overlay_qsv to allow: h264_qsv -> vpp_qsv=deinterlace,scale -> overlay_qsv but a53 captions were getting slightly garbled. I then tried a lot of combinations and a53 captions get garbled in all these: 1. QSV decode -> vpp_qsv -> overlay_qsv -> h264_qsv = bad closed captions 2. QSV decode -> overlay_qsv -> h264_qsv = bad closed captions 3. Software decode -> overlay_qsv -> h264_qsv = bad closed captions 4. Software decode -> overlay -> h264_qsv = bad closed captions I then decided apparently h264_qsv can mess up captions itself, but strangely not if using qsv decoding. I then settled on: vaapi_h264 -> deinterlace_vaapi -> scale_vaapi -> overaly_vaapi and a53 captions worked great and cpu usage (5-8%) was the lowest of any overlay I've seen, which means it's almost entirely in the gpu. Now I'm wondering if I should go back to using vaapi decoding and encoding for everything, I had started out this project wanting to use qsv since it in theory was the newest and had more features. Are there any downsides to using vaapi over qsv? I just want something that works and produces good quality transcoding. Thanks, Shane Warren

2 1

How to retain correct frame order when concatenating?
by Flopsy 30 Oct '24

30 Oct '24

Hi! I'm trying to concatenate these two single frame videos [https://files.catbox.moe/v85pqy.zip] but they always end up in the wrong order. I've tried with the following command: > ffmpeg -y -f concat -i filelist.txt -c copy outf.mkv ...where filelist.txt contains: > file '0012890a_9680.mkv' > file '0012890a_692070.mkv' But this gives me a 2 frame video with the frame order reversed. Examining the file with `ffprobe -show_frames`, it does indeed show the frames in reverse (some lines omitted): > [FRAME] > key_frame=1 > pts=42 > pts_time=0.042000 > best_effort_timestamp=42 > best_effort_timestamp_time=0.042000 > duration=41 > duration_time=0.041000 > pkt_pos=28240 > pkt_size=88803 > > [FRAME] > key_frame=1 > pts=0 > pts_time=0.000000 > best_effort_timestamp=0 > best_effort_timestamp_time=0.000000 > duration=41 > duration_time=0.041000 > pkt_pos=7976 > pkt_size=20246 It's interesting to note that, according to `pkt_pos`, the frames are written in correct order, and the PTS is also correct, however the order listed is somehow incorrect. If I reverse the order of the frames in filelist.txt, I get the order expected. So it seems like ffmpeg really wants to put the second frame first. I've also tried adding `-bsf:v setts=ts=N` to the ffmpeg command - the resulting PTS for the frames are 1 and 0, indicating correct packet order, but frames listed in reverse. I've also tried with mkvmerge, but get the same result (command: `mkvmerge --output outm.mkv 0012890a_9680.mkv + 0012890a_692070.mkv --append-to 1:0:0:0`) Does anyone know what's going on and how to get the frames in the correct order?

3 5

zscale and scale with HDR video
by Oliver Fromme 26 Oct '24

26 Oct '24

I've got a question for the experts ... So far I have successfully used zscale for scaling HDR video, because it is supposed to be better suited for HDR than the regular scale filter. However, I'm now in a situation where I need to change the scaling dynamically, depending on frame number -- that is, I have to use "n" (frame number) in the formula for width and/or height, along with the "eval=frame" option. Unfortunately, zscale does not support that, so I have to resort to using the regular scale filter. In the documentation I noticed that it supports options for setting the color space (in_color_matrix, out_color_matrix), so I set those to "bt2020" for HDR. Here's the complete command, broken up in POSIX shell syntax using variables for better readability: H='if(between(n\,47105\,47283)\,2160/(2032/1600)\,2160)' VF="scale=w=-1:h='$H':eval=frame:in_color_matrix=bt2020:out_color_matrix=bt2020" VF="${VF},pad=3840:2160:-1:-1:eval=frame" VF="${VF},crop=h=1600" HDR="hdr10=1:hdr10-opt=1:repeat-headers=1" HDR="${HDR}:colorprim=bt2020:transfer=smpte2084:colormatrix=bt2020nc" HDR="${HDR}:master-display=G(8500,39850)B(6550,2300)R(35400,14600)WP(15635,16450)L(10000000,0)" HDR="${HDR}:max-cll=0,0" ffmpeg -i input.mkv -vf "$VF" -map 0 -c copy -c:v libx265 \ -x265-params "$HDR" -pix_fmt yuv420p10le -preset slow -crf 20 \ -c:a ac3 -ac:a 6 -b:a 448k -default_mode infer_no_subs \ output.mkv Technically, that command works and runs without any errors or even warning messages. The resulting video looks "okay" on first sight. However ... when pausing the video and looking closer at the screen, I believe there are some small artefacts that are not in the source video. I'm not 100 % sure if my eyes deceive me or if it's real. And: If it's real, then is it caused by the scale filter? Or maybe do I just need to lower the CRF value, or maybe it's something else? And of course, I still remember the statement that the scale filter is not really recommended for HDR (zscale should be used instead, but I can't do that, as explained above). Can somebody give a recommendation for this situation? Is it okay to use the scale filter as in the above example? Can it be improved? Or -- is there a way to use the zscale filter with a dynamic formula that depends on frame number? Maybe I've overlooked something here. Best regards -- Oliver PS: Just in case somebody wonders why I'm doing that in the first place ... My source video is in 16:9 aspect ratio with 2160 pixel rows, but it contains black bands. I want to crop the video in order to get rid of the black bands. In the above example, the actual content is only 1600 pixel rows, but with one exception: There is a short section (just a few seconds) where the black bands are smaller. Within that section, the content occupies 2032 pixel rows. So, before cropping, I have to scale that section (only that section, not the rest!) down, so nothing gets lost when I crop to 1600 pixels. Additionally I have to insert a pad filter that adds black bands at the sides (left and right) within that section, so the width of the video stays constant.

2 1

How to resume ffmpeg vmaf evaluation from point in time
by tarun singhal 25 Oct '24

25 Oct '24

Hi Team, I have 3-hour long video on which I run VMAF evaluation against its transcoded version. and generally, it takes 5-6 hr to finish. I am running these on kubernetes pods. I have a continuous deployment pipeline. What happens is often while vmaf is about to finish someone merges code and it kills the old pod in which case I end up restarting the entire process again. Is there a way in ffmpeg world where I can keep a track of how much vmaf is done and resume from a point in time? Thanks Tarun

2 1