- Cin - Cinelerra-GG Mailing Lists

CinGG Analog S-Video to Digital HDMI converter + ms2130 UHD HDMI-USB3 Capture card
by Terje J. Hanssen 05 May '25

05 May '25

https://www.mail-archive.com/cin@lists.cinelerra-gg.org/msg05748.html I contine this new thread with an short extract from the long growing [Cin] ms2130 HDMI-USB3 UHD capture card - not quite as expected? But shouldn't a 422 and not a 420 output format be created? >> it seems in some cases ffmpeg not geared towards preserving most information, just picks up most commonly defined/first pixel format. >> Try mpeg2video as output codec, and/or specify pixel format (pix_fmt) more explicitly before output's filename? >>> Could you possibly examplify the latter? >>>> I think I just theorized ffmpeg may try to get some info from alsa device file directly, as it tries with /dev/video. But I never >>>> tried this way, so I can be totally wrong! ------------------------------- Instead of 50i (pal interlaced fields), I try 50 fps (progressive) that at least is a pal frequence. And yes, it seems from the output that ffmpeg uses mpeg1video as default output codec. Furter I try to combine the current line ffmpeg -hide_banner -f v4l2 -input_format yuyv422 -video_size 720x576 -framerate 25 -t 20 -i /dev/video2 out-720x576_ms2130.mpeg with and utilize the syntax developed 2021 for "Blu-ray compliant MPEG-2 SD video with LPCM audio" ffmpeg -i dv28.dv -c:v mpeg2video -refs 1 -bf 2 -b:v 25M -maxrate 25M -minrate 25M -bufsize 45M -muxrate 45M -dc 10 -c:a pcm_s16be -f vob dv28.mpg and for the current ongoing "DVD compliant MPEG-2 Video and PCM Audio" ffmpeg -hide_banner -i dv01.dv -c:v mpeg2video -refs 1 -bf 2 -b:v 8M -maxrate 8M -minrate 8M -bufsize 20M -muxrate 11M -dc 10 -c:a pcm_dvd -f vob -flags +ilme+ildct dv01.mpg to first ffmpeg -hide_banner -f v4l2 -input_format yuyv422 -video_size 720x576 -framerate 50 -c:v mpeg2video -refs 1 -bf 2 -b:v 8M -maxrate 8M -minrate 8M -bufsize 20M -muxrate 11M -dc 10 -c:a pcm_dvd -f vob -t 20 -i /dev/video2 720x576_50p_ms2130.mpg Option b:v (video bitrate (please use -b:v)) cannot be applied to input url /dev/video2 -- you are trying to apply an input option to an output file or vice versa. Move this option before the file it belongs to. Error parsing options for input file /dev/video2. Error opening input files: Invalid argument and next revised with no output. That is I need help to debug the syntax error: ffmpeg -hide_banner -f v4l2 -input_format yuyv422 -video_size 720x576 -framerate 50 -i /dev/video2 -c:v mpeg2video -refs 1 -bf 2 -b:v 8M -maxrate 8M -minrate 8M -bufsize 20M -muxrate 11M -dc 10 -c:a pcm_dvd -f vob -t 20 720x576_50p_ms2130.mpg [video4linux2,v4l2 @ 0x55db770ef9c0] Dequeued v4l2 buffer contains corrupted data (0 bytes). Input #0, video4linux2,v4l2, from '/dev/video2': Duration: N/A, start: 0.000000, bitrate: 331776 kb/s Stream #0:0: Video: rawvideo (YUY2 / 0x32595559), yuyv422, 720x576, 331776 kb/s, 50 fps, 50 tbr, 1000k tbn Stream mapping: Stream #0:0 -> #0:0 (rawvideo (native) -> mpeg2video (native)) Press [q] to stop, [?] for help [video4linux2,v4l2 @ 0x55db770ef9c0] Dequeued v4l2 buffer contains corrupted data (0 bytes). Last message repeated 30 times [mpeg2video @ 0x55db77114f00] Warning vbv_delay will be set to 0xFFFF (=VBR) as the specified vbv buffer is too large for the given bitrate! Output #0, vob, to '720x576_50p_ms2130.mpg': Metadata: encoder : Lavf59.27.100 Stream #0:0: Video: mpeg2video (4:2:2), yuv422p(tv, progressive), 720x576, q=2-31, 8000 kb/s, 50 fps, 90k tbn Metadata: encoder : Lavc59.37.100 mpeg2video Side data: cpb: bitrate max/min/avg: 8000000/8000000/8000000 buffer size: 20000000 vbv_delay: N/A frame= 0 fps=0.0 q=0.0 Lsize= 0kB time=00:00:00.00 bitrate=N/A speed= 0x video:0kB audio:0kB subtitle:0kB other streams:0kB global headers:0kB muxing overhead: unknown Output file is empty, nothing was encoded (check -ss / -t / -frames parameters if used)

2 9

New Samsung codec "APV"
by Andrea paz 01 May '25

01 May '25

Ffmpeg has introduced (in decoding) a new “mezzanine codec” from Samsung: apv. There is also a proposal for an “OpenAPV” codec. The news is not particularly interesting: CinGG will read it because ffmpeg will read it; but the codec “draft” is useful and thorough in explaining how a codec is built and how it works. It can be found here: https://datatracker.ietf.org/doc/draft-lim-apv/ PS: I also attach an image of the “quality pentagon ” of a video. It is OT, but I particularly like it for its clarity and simplicity.

2 1

Updated AppImages for April 2025
by Phyllis Smith 30 Apr '25

30 Apr '25

There are now updated AppImages, Release Notes, and PDF/HTML manuals on the website and Andrey's packages are up to date at: https://github.com/einhander/cin-gg-packages/releases Release notes included below with some concern for the Speed mod as noted with no real concern for Blur and Titler. *GIT for Cinelerra-GG has the following changes from 04/01/2025-04/302025* *Blend Algebra and Blend Program plugins:* An additional SYSLIB Blend Program of tone_mapping.bp has been created by *Andrea* for tonemapping an HDR media and report its values within the range (0, 1.0). For usage, reference Example 6 in the file at: https://cinelerra-gg.org/download/testing/BlendPluginExamples/Examples.txt For netBSD, modifications were added by Andrew to fix shebang-perl and to not use ldl in Makefile. *Other small changes include:* 4.0 cmake requires 3 patches that *Andrew* has provided and Andrea had discovered/tested on Arch for x265 and libsvtav1. A speed bug has been found and was described reproducibly by *Lukas*. A one line change to auto.C seems to fix it, but it is unknown if reverting that line will lead to a previously diagnosed problem reoccuring. The old version of ffnvcodec (Video codec SDK) which was 10.0.26.0 was added back to the source as ffnvcodec-old.tar.xz to accommodate versions before the updated version of ffnvcodec for 12.2.72.0 was installed as the default. Blur and Titler plugins were modified so that these plugins can be used in combination with others (e.g. Blue Banana) without introducing the clipping into the process. Andrew provided the mod and Andrea tested it to ensure it worked without clipping and IgorB tested to make sure there was no impact to current usage of Blur and the Titler. Tiff was updated from 4.6.0t to 4.7.0.

1 0

Tried ffv1_vulkan ....
by Andrew Randrianasulu 29 Apr '25

29 Apr '25

It looks like tonemapping ate encoder's lunch ;) RADV_PERFTEST=video_decode,video_encode time -p ./ffmpeg -init_hw_device vulkan=vulkan -filter_hw_device vulkan -hwaccel vulkan -i ~/K38_sdcard1/Documents/iPhone11_4K-recorder_59.940HDR10.mov -vf libplacebo=format=yuv444p:w=1920:h=1080:force_original_aspect_ratio=decrease:normalize_sar=true:upscaler=ewa_lanczos:downscaler=ewa_lanczos:colorspace=bt709:color_primaries=bt709:color_trc=bt709:range=tv -c:a copy -c:v ffv1_vulkan -slicecrc 0 -level 4 -strict -2 -async_depth 5 -slices_v 32 -slices_h 32 -f matroska -benchmark /dev/null frame= 1148 fps=4.6 q=-0.0 Lsize= 2302672KiB time=00:00:19.16 bitrate=984419.6kbits/s speed=0.0769x bench: utime=28.586s stime=20.721s rtime=249.150s bench: maxrss=587076KiB real 251.00 user 28.67 sys 20.79 ===== But may be faster GPU/pcie bus like in Andrea's machine will speed it up enough to be practical

3 5

xavc-intra class 480
by Andrew Randrianasulu 28 Apr '25

28 Apr '25

In attempt at speeding up hdr10 4k HEVC video playback I tried to re-encode it into AVC-intra using this guide: https://www.voukoder.org/forum/thread/1163-replicate-sony-s-xavc-i-s-format/ basically either ffmpeg -i input -c:v libx264 -avcintra-class 480 -pix_fmt yuv422p10le -preset ultrafast -x264opts avcintra-flavor=sony -c:a pcm_s24le -ar 48k output.mxf or for half size ffmpeg -i ~/K38_sdcard1/Documents/iPhone11_4K-recorder_59.940HDR10.mov -s 1920x1080 -c:v libx264 -avcintra-class 200 -pix_fmt yuv422p10le -preset ultrafast -x264opts avcintra-flavor=sony -c:a pcm_s24le -ar 48k output-half.mxf Note, line errors out when I tried class 480 for 1920x1080 frame size Both files load into cingg Original file (4k hdr hevc, 59.94 fps) plays at around 22 fps at output x11, project rgba-float. But seeking is sad. Not to mention lack of tonemapping. First 4k avc-intra 4k reencode plays at around 33 fps and seeks much better (no artefacts) 1920x1080 proxy plays at 60 fps, and if I check play every frame and try 2x speed forward - at 83 Fps. But trying two FullHD tracks referencing same "proxy" file set to different fade leves with project set to RGBA-float kills fps back to 2-3 :) Encoding at ultrafast preset was 4.5 fps for 4k version and 12 fps for FullHD version. Filesizes: du -h *.mxf 1.0G output-half.mxf 2.2G output.mxf this is just for frame= 1149 time=00:00:19.15 Well, I guess without some serious GPU upgrade this machine not gonna play HDR 4k :) Regular 8 bit/SDR 4k somewhat plays in firefox (if file just av1, 24 fps), at least on Linux where I can set GPU frequency a bit higher than default on nouveau/libre drivers.

3 6

New speed winner for hw tonemapping!
by Andrew Randrianasulu 28 Apr '25

28 Apr '25

RADV_PERFTEST=video_decode,video_encode time -p ./ffmpeg -init_hw_device vulkan=vulkan -filter_hw_device vulkan -hwaccel vulkan -i ~/K38_sdcard1/Documents/iPhone11_4K-recorder_59.940HDR10.mov -vf libplacebo=w=1920:h=1080:force_original_aspect_ratio=decrease:normalize_sar=true:upscaler=ewa_lanczos:downscaler=ewa_lanczos:colorspace=bt709:color_primaries=bt709:color_trc=bt709:range=tv -c:a copy -c:v libx264 -f mp4 -benchmark /dev/shm/ffmpeg-git-libplacebo-vulkan-2k.mp4 frame= 1148 fps= 10 q=-1.0 Lsize= 30435KiB time=00:00:19.13 bitrate=13029.3kbits/s speed=0.167x bench: utime=355.600s stime=25.878s rtime=114.342s so nearly 10 fps with scaled to FHD input! (otherwise 32bit x264 OOMs) ffmpeg git 8bb682d454990a1049a21f1f51442205ea3337e9 configured as ./configure --enable-opencl --disable-debug --enable-libx265 --enable-libx264 --enable-gpl --enable-libplacebo --enable-vulkan --enable-libshaderc --enable-libzimg --enable-libaom --enable-libdav1d --enable-libsoxr --enable-libfontconfig --enable-libfreetype --enable-libfribidi --enable-gnutls --enable-libass --enable-libbluray --enable-libcdio --enable-frei0r --enable-libgsm --enable-openal --enable-libopus --enable-librtmp --enable-libsnappy --enable-libspeex --enable-libssh --enable-libtheora --enable-libtwolame --enable-libv4l2 --enable-libvidstab --enable-libvorbis --enable-libvpx --enable-libwebp Cpu usage for just decoding also was lower than in vaapi case, so *some* use for Vulkan decode, contrary to my initial sceptecism!

3 7

I tried to make sws_scale go multithread ..
by Andrew Randrianasulu 28 Apr '25

28 Apr '25

BUT FAILED. in sense my mod apparently just leaks (swsContext?) so it can't play even 20 sec of 1440*1080 video to completion in termux, running on tablet with 3Gb ram. There seems to be difference if I set thread count manually (on file reading only for now) but it looks negative, in other words higher thread count -> slower total playback .. tried 8, 4,1,16 on 8 cpu tablet. Speed was going down from 18 fps for unaltered cingg to 13 with 8 threads and crash/oom with 16 ;) And if I try to use sws_scale_frame I just got all blank/black frames on timeline .... I think switching to sws_scale_frame is important if we want to try new colormanagement support in ffmpeg.git ... Resources used, apart from that Blender diff: https://github.com/mltframework/mlt/blob/master/src/modules/avformat/filter…

1 0

threaded swscale in blender
by Andrew Randrianasulu 27 Apr '25

27 Apr '25

Interesting diff, may be cingg can use something like this for speeding up swscale conversions? https://projects.blender.org/blender/blender/pulls/116309/commits/4647c80f5… But it seems that only upcoming ffmpeg version will allow to set transfer characteristics directly on avframe during scaling. So may be until this day our colorspace plugin should grow some knowledge about various gamma systems and also white points?

1 1

Can sws_scale() set trc?
by Andrew Randrianasulu 27 Apr '25

27 Apr '25

Or we need to go and set in->color_trc out->color_trc on avframes directly, like libavfilter/vf_scale.c does ?

1 0

libavutil/csp.h reading
by Andrew Randrianasulu 27 Apr '25

27 Apr '25

=== /** * Returns the ITU EOTF corresponding to a given TRC. This converts from the * signal level [0,1] to the raw output display luminance in nits (cd/m^2). * This is done per channel in RGB space, except for AVCOL_TRC_SMPTE428, which * assumes CIE XYZ in- and output. * * @return A pointer to the function implementing the given TRC, or NULL if no * such function is defined. * * @note In general, the resulting function is defined (wherever possible) for * out-of-range values, even though these values do not have a physical * meaning on the given display. Users should clamp inputs (or outputs) * if this behavior is not desired. * * This is also the case for functions like PQ, which are defined over an * absolute signal range independent of the target display capabilities. */ av_csp_eotf_function av_csp_itu_eotf(enum AVColorTransferCharacteristic trc); I guess this can be used for potential HDR waveform scope display?

1 0