Arch Linux User Repository

Please, this package follow Arch Linux ffmpeg package version. Don't put ffmpeg-obs out-of-date if both packages are on the same version.

OBS Project actually patch FFmpeg to fix some issues and also add RIST support starting from 27.2.

So I created this package, a ffmpeg package that ensure that you have those.

I really wanted to not do this but some fixes are needed, so I did my best to add those "feature-set options".

They allow you to build ffmpeg with feature-set based on other ffmpeg packages from AUR.

Read the PKGBUILD for more information about them.

Building with ffnvcodec-headers 12.2 is fixed.

Currently, this is failing to build due to the most recent update to the official ffnvcodec-headers package (the official arch package updated about 16 hours ago as of this comment):

libavcodec/nvenc.c: In function ‘nvenc_setup_hevc_config’:
libavcodec/nvenc.c:1373:9: error: ‘NV_ENC_CONFIG_HEVC’ has no member named ‘pixelBitDepthMinus8’
 1373 |     hevc->pixelBitDepthMinus8 = IS_10BIT(ctx->data_pix_fmt) ? 2 : 0;
      |         ^~
libavcodec/nvenc.c: In function ‘nvenc_setup_av1_config’:
libavcodec/nvenc.c:1458:8: error: ‘NV_ENC_CONFIG_AV1’ has no member named ‘inputPixelBitDepthMinus8’
 1458 |     av1->inputPixelBitDepthMinus8 = IS_10BIT(ctx->data_pix_fmt) ? 2 : 0;
      |        ^~
libavcodec/nvenc.c:1459:8: error: ‘NV_ENC_CONFIG_AV1’ has no member named ‘pixelBitDepthMinus8’
 1459 |     av1->pixelBitDepthMinus8 = (IS_10BIT(ctx->data_pix_fmt) || ctx->highbitdepth) ? 2 : 0;
      |        ^~
CC  libavcodec/opus_celt.o
libavcodec/nvenc.c: In function ‘nvenc_map_buffer_format’:
libavcodec/nvenc.c:1692:16: error: ‘NV_ENC_BUFFER_FORMAT_YV12_PL’ undeclared (first use in this function); did you mean ‘NV_ENC_BUFFER_FORMAT_YV12’?
 1692 |         return NV_ENC_BUFFER_FORMAT_YV12_PL;
      |                ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
      |                NV_ENC_BUFFER_FORMAT_YV12
libavcodec/nvenc.c:1692:16: note: each undeclared identifier is reported only once for each function it appears in
libavcodec/nvenc.c:1694:16: error: ‘NV_ENC_BUFFER_FORMAT_NV12_PL’ undeclared (first use in this function); did you mean ‘NV_ENC_BUFFER_FORMAT_NV12’?
 1694 |         return NV_ENC_BUFFER_FORMAT_NV12_PL;
      |                ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
      |                NV_ENC_BUFFER_FORMAT_NV12
CC  libavcodec/opus_metadata_bsf.o
libavcodec/nvenc.c:1700:16: error: ‘NV_ENC_BUFFER_FORMAT_YUV444_PL’ undeclared (first use in this function); did you mean ‘NV_ENC_BUFFER_FORMAT_YUV444’?
 1700 |         return NV_ENC_BUFFER_FORMAT_YUV444_PL;
      |                ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
      |                NV_ENC_BUFFER_FORMAT_YUV444

It appeears that about 6 days ago the official Videolan.org project (Official git.videolan.org Commit Diff here) removed the pixelBitDepthMinus8 property from the NV_ENC_CONFIG_HEVC struct:

@@ -1793,7 +1843,7 @@ typedef struct _NV_ENC_CONFIG_HEVC
     uint32_t repeatSPSPPS                          :1;              /**< [in]: Set 1 to output VPS,SPS and PPS for every IDR frame.*/
     uint32_t enableIntraRefresh                    :1;              /**< [in]: Set 1 to enable gradual decoder refresh or intra refresh. If the GOP structure uses B frames this will be ignored */
     uint32_t chromaFormatIDC                       :2;              /**< [in]: Specifies the chroma format. Should be set to 1 for yuv420 input, 3 for yuv444 input.*/
-    uint32_t pixelBitDepthMinus8                   :3;              /**< [in]: Specifies pixel bit depth minus 8. Should be set to 0 for 8 bit input, 2 for 10 bit input.*/
+    uint32_t reserved3                             :3;              /**< [in]: Reserved and must be set to 0.*/
     uint32_t enableFillerDataInsertion             :1;              /**< [in]: Set to 1 to enable insertion of filler data in the bitstream.
                                                                                This flag will take effect only when CBR rate control mode is in use and both
                                                                                NV_ENC_INITIALIZE_PARAMS::frameRateNum and

As well as removed the other two properties inputPixelBitDepthMinus8 and pixelBitDepthMinus8 from the NV_ENC_CONFIG_AV1 struct:

@@ -1910,10 +1973,7 @@ typedef struct _NV_ENC_CONFIG_AV1
     uint32_t enableBitstreamPadding         : 1;                    /**< [in]: Set 1 to enable bitstream padding. */
     uint32_t enableCustomTileConfig         : 1;                    /**< [in]: Set 1 to enable custom tile configuration: numTileColumns and numTileRows must have non zero values and tileWidths and tileHeights must point to a valid address  */
     uint32_t enableFilmGrainParams          : 1;                    /**< [in]: Set 1 to enable custom film grain parameters: filmGrainParams must point to a valid address  */
-    uint32_t inputPixelBitDepthMinus8       : 3;                    /**< [in]: Specifies pixel bit depth minus 8 of video input. Should be set to 0 for 8 bit input, 2 for 10 bit input.*/
-    uint32_t pixelBitDepthMinus8            : 3;                    /**< [in]: Specifies pixel bit depth minus 8 of encoded video. Should be set to 0 for 8 bit, 2 for 10 bit.
-                                                                               HW will do the bitdepth conversion internally from inputPixelBitDepthMinus8 -> pixelBitDepthMinus8 if bit dpeths differ
-                                                                               Support for 8 bit input to 10 bit encode conversion only */
+    uint32_t reserved4                      : 6;                    /**< [in]: Reserved and must be set to 0.*/
     uint32_t reserved                       : 14;                   /**< [in]: Reserved bitfields.*/
     uint32_t idrPeriod;                                             /**< [in]: Specifies the IDR/Key frame interval. If not set, this is made equal to gopLength in NV_ENC_CONFIG.Low latency application client can set IDR interval to NVENC_INFINITE_GOPLENGTH so that IDR frames are not inserted automatically. */
     uint32_t intraRefreshPeriod;                                    /**< [in]: Specifies the interval between successive intra refresh if enableIntrarefresh is set. Requires enableIntraRefresh to be set.

Sadly, I'm not smart enough to figure out what to replace with since they just flat out removed those two properties. Maybe this is a bug (nervously hopeful)?

Anybody else stuck with being unable to build here's what I did to re-build:

Downgraded my official archlinux ffnvcodec-headers package by cloning it from: https://gitlab.archlinux.org/archlinux/packaging/packages/ffnvcodec-headers

Then I just checked out the previous 12.1.14.0-1 tagged commit:

git clone https://gitlab.archlinux.org/archlinux/packaging/packages/ffnvcodec-headers.git
git checkout -b 12.1.14.0-1 tags/12.1.14.0-1
makepkg -s
sudo pacman -U ffnvcodec-headers-12.1.14.0-1-any.pkg.tar.zst 

So far so good.. Though I need to test OBS with hardware NVENC... Though my GPU is not powerful enough to do HEVC or AV1...

Thanks, that did the trick.

pod2man from the perl package is not present in your $PATH.

17:43:24 MAN    doc/ffmpeg.1
17:43:24 /bin/sh: line 1: pod2man: command not found
17:43:24 make: *** [doc/Makefile:98: doc/ffmpeg.1] Error 127
17:43:24 make: *** Waiting for unfinished jobs....
17:43:24 MAN    doc/ffplay.1
17:43:24 /bin/sh: line 1: pod2man: command not found
17:43:24 make: *** [doc/Makefile:98: doc/ffplay.1] Error 127
17:43:24 ==> ERROR: A failure occurred in build().
17:43:24     Aborting...

https://man.archlinux.org/man/core/perl/pod2man.1perl.en suggests it's a part of the core/perl package, which is in fact installed. Might be a local system configuration issue.

i use paru, and it did rebulid vlc-luajit, i meant that the problem is some sort of ac3 codec error which got fixed when i switched to normal ffmpeg

but mpv was actually libplacebo error

VLC packages (vlc and vlc-luajit) no longer enable libplacebo since a while.

If I have to guess, you didn't try to clean re-build ffmpeg-obs since libplacebo package got its new soname.

This package breaks vlc and mpv now, with a libplacebo.so.264 missing error, I really want to use obs-studio-tytan, cant you just embed ffmpeg into the package itself or something?

@kescherAUR, sorry for the delay I discovered that I messed up the FFmpeg version commit hash.

6.1.1-3 should be good, sorry for everyone that met issue with earlier 6.1.1 packages.

With FFMPEG_OBS_DECKLINK, FFMPEG_OBS_LIBFDK_AAC, FFMPEG_OBS_SVT and FFMPEG_OBS_VULKAN enabled, this happens:

patching file libavformat/avformat.h
Hunk #1 succeeded at 1030 with fuzz 2 (offset -98 lines).
patching file libavformat/utils.c
Hunk #1 succeeded at 56 (offset 1 line).
patching file libavformat/flvdec.c
Hunk #1 succeeded at 1108 (offset 31 lines).
patching file libavcodec/libaomenc.c
Hunk #1 succeeded at 1460 (offset 123 lines).
patching file configure
Hunk #1 FAILED at 291.
Hunk #2 succeeded at 1893 (offset 41 lines).
Hunk #3 succeeded at 3471 (offset 66 lines).
Hunk #4 succeeded at 6830 (offset 150 lines).
1 out of 4 hunks FAILED -- saving rejects to file configure.rej
patching file libavcodec/Makefile
Hunk #1 succeeded at 1147 with fuzz 1 (offset 21 lines).
patching file libavcodec/allcodecs.c
Hunk #1 FAILED at 812.
1 out of 1 hunk FAILED -- saving rejects to file libavcodec/allcodecs.c.rej
patching file libavcodec/libsvt_hevc.c