【分享】使用ffmpeg5.0转码杜比视界Dolby Vision(Profile5) 片源(含播放教程)

[nilaoda]

前言

如 #52 所述，ffmpeg5.0通过整合libplacebo正式支持了DoVi(杜比视界)的映射和混流. 本文尝试使用FFmpeg转码或播放杜比视界Profile5片源.

libplacebo滤镜

>>ffmpeg -h filter=libplacebo

Filter libplacebo
  Apply various GPU filters from libplacebo
    Inputs:
       #0: default (video)
    Outputs:
       #0: default (video)
libplacebo AVOptions:
   w                 <string>     ..FV....... Output video width (default "iw")
   h                 <string>     ..FV....... Output video height (default "ih")
   format            <string>     ..FV....... Output video format
   force_original_aspect_ratio <int>        ..FV....... decrease or increase w/h if necessary to keep the original AR (from 0 to 2) (default disable)
     disable         0            ..FV.......
     decrease        1            ..FV.......
     increase        2            ..FV.......
   force_divisible_by <int>        ..FV....... enforce that the output resolution is divisible by a defined integer when force_original_aspect_ratio is used (from 1 to 256) (default 1)
   normalize_sar     <boolean>    ..FV....... force SAR normalization to 1:1 (default true)
   pad_crop_ratio    <float>      ..FV.....T. ratio between padding and cropping when normalizing SAR (0=pad, 1=crop) (from 0 to 1) (default 0)
   colorspace        <int>        ..FV.....T. select colorspace (from -1 to 14) (default auto)
     auto            -1           ..FV....... keep the same colorspace
     gbr             0            ..FV.......
     bt709           1            ..FV.......
     unknown         2            ..FV.......
     bt470bg         5            ..FV.......
     smpte170m       6            ..FV.......
     smpte240m       7            ..FV.......
     ycgco           8            ..FV.......
     bt2020nc        9            ..FV.......
     bt2020c         10           ..FV.......
     ictcp           14           ..FV.......
   range             <int>        ..FV.....T. select color range (from -1 to 2) (default auto)
     auto            -1           ..FV....... keep the same color range
     unspecified     0            ..FV.......
     unknown         0            ..FV.......
     limited         1            ..FV.......
     tv              1            ..FV.......
     mpeg            1            ..FV.......
     full            2            ..FV.......
     pc              2            ..FV.......
     jpeg            2            ..FV.......
   color_primaries   <int>        ..FV.....T. select color primaries (from -1 to 22) (default auto)
     auto            -1           ..FV....... keep the same color primaries
     bt709           1            ..FV.......
     unknown         2            ..FV.......
     bt470m          4            ..FV.......
     bt470bg         5            ..FV.......
     smpte170m       6            ..FV.......
     smpte240m       7            ..FV.......
     film            8            ..FV.......
     bt2020          9            ..FV.......
     smpte428        10           ..FV.......
     smpte431        11           ..FV.......
     smpte432        12           ..FV.......
     jedec-p22       22           ..FV.......
     ebu3213         22           ..FV.......
   color_trc         <int>        ..FV.....T. select color transfer (from -1 to 18) (default auto)
     auto            -1           ..FV....... keep the same color transfer
     bt709           1            ..FV.......
     unknown         2            ..FV.......
     bt470m          4            ..FV.......
     bt470bg         5            ..FV.......
     smpte170m       6            ..FV.......
     smpte240m       7            ..FV.......
     linear          8            ..FV.......
     iec61966-2-4    11           ..FV.......
     bt1361e         12           ..FV.......
     iec61966-2-1    13           ..FV.......
     bt2020-10       14           ..FV.......
     bt2020-12       15           ..FV.......
     smpte2084       16           ..FV.......
     arib-std-b67    18           ..FV.......
   upscaler          <string>     ..FV.....T. Upscaler function (default "spline36")
   downscaler        <string>     ..FV.....T. Downscaler function (default "mitchell")
   lut_entries       <int>        ..FV.....T. Number of scaler LUT entries (from 0 to 256) (default 0)
   antiringing       <float>      ..FV.....T. Antiringing strength (for non-EWA filters) (from 0 to 1) (default 0)
   sigmoid           <boolean>    ..FV.....T. Enable sigmoid upscaling (default true)
   apply_filmgrain   <boolean>    ..FV.....T. Apply film grain metadata (default true)
   apply_dolbyvision <boolean>    ..FV.....T. Apply Dolby Vision metadata (default true)
   deband            <boolean>    ..FV.....T. Enable debanding (default false)
   deband_iterations <int>        ..FV.....T. Deband iterations (from 0 to 16) (default 1)
   deband_threshold  <float>      ..FV.....T. Deband threshold (from 0 to 1024) (default 4)
   deband_radius     <float>      ..FV.....T. Deband radius (from 0 to 1024) (default 16)
   deband_grain      <float>      ..FV.....T. Deband grain (from 0 to 1024) (default 6)
   brightness        <float>      ..FV.....T. Brightness boost (from -1 to 1) (default 0)
   contrast          <float>      ..FV.....T. Contrast gain (from 0 to 16) (default 1)
   saturation        <float>      ..FV.....T. Saturation gain (from 0 to 16) (default 1)
   hue               <float>      ..FV.....T. Hue shift (from -3.14159 to 3.14159) (default 0)
   gamma             <float>      ..FV.....T. Gamma adjustment (from 0 to 16) (default 1)
   peak_detect       <boolean>    ..FV.....T. Enable dynamic peak detection for HDR tone-mapping (default true)
   smoothing_period  <float>      ..FV.....T. Peak detection smoothing period (from 0 to 1000) (default 100)
   minimum_peak      <float>      ..FV.....T. Peak detection minimum peak (from 0 to 100) (default 1)
   scene_threshold_low <float>      ..FV.....T. Scene change low threshold (from -1 to 100) (default 5.5)
   scene_threshold_high <float>      ..FV.....T. Scene change high threshold (from -1 to 100) (default 10)
   overshoot         <float>      ..FV.....T. Tone-mapping overshoot margin (from 0 to 1) (default 0.05)
   intent            <int>        ..FV.....T. Rendering intent (from 0 to 3) (default relative)
     perceptual      0            ..FV....... Perceptual
     relative        1            ..FV....... Relative colorimetric
     absolute        3            ..FV....... Absolute colorimetric
     saturation      2            ..FV....... Saturation mapping
   tonemapping       <int>        ..FV.....T. Tone-mapping algorithm (from 0 to 6) (default bt.2390)
     clip            0            ..FV....... Hard-clipping
     mobius          1            ..FV....... Mobius tone-mapping
     reinhard        2            ..FV....... Reinhard tone-mapping
     hable           3            ..FV....... Hable/Filmic tone-mapping
     gamma           4            ..FV....... Gamma tone-mapping
     linear          5            ..FV....... Linear tone-mapping
     bt.2390         6            ..FV....... ITU-R BT.2390 tone-mapping
   tonemapping_param <float>      ..FV.....T. Tunable parameter for some tone-mapping functions (from 0 to 100) (default 0)
   desaturation_strength <float>      ..FV.....T. Desaturation strength (from 0 to 1) (default 0.9)
   desaturation_exponent <float>      ..FV.....T. Desaturation exponent (from 0 to 10) (default 0.2)
   desaturation_base <float>      ..FV.....T. Desaturation base (from 0 to 10) (default 0.18)
   max_boost         <float>      ..FV.....T. Tone-mapping maximum boost (from 1 to 10) (default 1)
   gamut_warning     <boolean>    ..FV.....T. Highlight out-of-gamut colors (default false)
   gamut_clipping    <boolean>    ..FV.....T. Enable colorimetric gamut clipping (default true)
   dithering         <int>        ..FV.....T. Dither method to use (from -1 to 3) (default blue)
     none            -1           ..FV....... Disable dithering
     blue            0            ..FV....... Blue noise
     ordered         1            ..FV....... Ordered LUT
     ordered_fixed   2            ..FV....... Fixed function ordered
     white           3            ..FV....... White noise
   dither_lut_size   <int>        ..FV....... Dithering LUT size (from 1 to 8) (default 6)
   dither_temporal   <boolean>    ..FV.....T. Enable temporal dithering (default false)
   cones             <flags>      ..FV.....T. Colorblindness adaptation model (default 0)
     l                            ..FV....... L cone
     m                            ..FV....... M cone
     s                            ..FV....... S cone
   cone-strength     <float>      ..FV.....T. Colorblindness adaptation strength (from 0 to 10) (default 0)
   custom_shader_path <string>     ..FV....... Path to custom user shader (mpv .hook format)
   custom_shader_bin <binary>     ..FV....... Custom user shader as binary (mpv .hook format)
   skip_aa           <boolean>    ..FV.....T. Skip anti-aliasing (default false)
   polar_cutoff      <float>      ..FV.....T. Polar LUT cutoff (from 0 to 1) (default 0)
   disable_linear    <boolean>    ..FV.....T. Disable linear scaling (default false)
   disable_builtin   <boolean>    ..FV.....T. Disable built-in scalers (default false)
   force_icc_lut     <boolean>    ..FV.....T. Force the use of a full ICC 3DLUT for color mapping (default false)
   force_dither      <boolean>    ..FV.....T. Force dithering (default false)
   disable_fbos      <boolean>    ..FV.....T. Force-disable FBOs (default false)

tonemapping

利用Vulkan支持，将DoVi(P5,ICtCp)片源映射到普通的色彩空间：

• 默认

产生BT.2020-PQ的10bit HEVC视频:

ffmpeg -y -init_hw_device vulkan=vulkan -filter_hw_device vulkan -i "Dolby Vsion Amaze.mp4" -vf "hwupload,libplacebo=iw:ih:yuv420p10le,hwdownload,format=yuv420p10le" -c:v libx265 -c:a copy "output.mp4"

(iw, ih为输入视频的大小，也可以手动指定, 例如1920:1080)

• 指定色彩空间

产生BT.709的10bit HEVC视频:

ffmpeg -y -init_hw_device vulkan=vulkan -filter_hw_device vulkan -i "Dolby Vsion Amaze.mp4" -vf "hwupload,libplacebo=iw:ih:format=yuv420p10le:colorspace=bt709:color_primaries=bt709:color_trc=bt709,hwdownload,format=yuv420p10le" -c:v libx265 -c:a copy "output709.mp4"

产生BT.2020-HLG的10bit HEVC视频:

ffmpeg -y -init_hw_device vulkan=vulkan -filter_hw_device vulkan -i "Dolby Vsion Amaze.mp4" -vf "hwupload,libplacebo=iw:ih:format=yuv420p10le:colorspace=bt2020nc:color_primaries=bt2020:color_trc=arib-std-b67,hwdownload,format=yuv420p10le" -c:v libx265 -c:a copy "outputHLG.mp4"

• CUDA硬件加速

Decode & Encode:

ffmpeg -y -hwaccel cuda -hwaccel_output_format cuda -extra_hw_frames 3 -i "Dolby Vsion Amaze.mp4" -vf "hwupload=derive_device=vulkan,libplacebo=iw:ih:format=yuv420p10le:colorspace=bt2020nc:color_primaries=bt2020:color_trc=arib-std-b67,hwdownload,format=yuv420p10le" -c:v hevc_nvenc -c:a copy "outputHLG.mp4"

播放

• pipe到ffplay

ffmpeg -y -hwaccel cuda -hwaccel_output_format cuda -extra_hw_frames 3 -i "Dolby Vsion Amaze.mp4" -vf "hwupload=derive_device=vulkan,libplacebo=1280:720:format=yuv420p10le:colorspace=bt2020nc:color_primaries=bt2020:color_trc=arib-std-b67,hwdownload,format=yuv420p10le" -c:v hevc_nvenc -c:a copy -f mpegts - | ffplay - 

• 使用支持的开源播放器

今年早些时候，MPV 已随着ffmpeg5.0发布测试性地支持了杜比视界的tonemapping，只需要加入--vo=gpu-next参数

mpv --vo=gpu-next "C:\Downloads\Dolby Vsion Amaze.mp4"

拓展阅读

• https://newbedev.com/how-to-use-gpu-to-accelerate-the-processing-speed-of-ffmpeg-filter
• https://trac.ffmpeg.org/wiki/HWAccelIntro

[bone-cancer]

您好，我用您提供的产生BT.709的10bit HEVC视频的命令，试图将profile5的视频转码至BT709,但是报错Cannot allocate all planes in a single allocation,device requires dedicated image allcation，Error while filtering:cannot allocate memory 请问您有遇到过这个问题吗

[nilaoda]

升级显卡驱动后重启电脑试试

[bone-cancer]

升级驱动后成功了！感谢大佬！

[shan8901]

我用软件编码一切正常，但用您给出的CUDA加速转码命令，会报错：

Impossible to convert between the formats supported by the filter 'Parsed_hwupload_0' and the filter 'auto_scale_0'
Error reinitializing filters!
Failed to inject frame into filter network: Function not implemented

CUDA驱动已经是最新的，ffmpeg版本是6.0

请教是什么问题引起的？

[hym3242]

这个 auto_scale_0 不应该被自动插入至 filtergraph, 这个现象只会在前一个 filter 支持的输出格式和后一个 filter 支持的输入格式没有交集时出现，建议确认你的 hwcontext 是 vulkan(libplacebo 不支持 CUDA 像素格式), 且不要在 hwupload 和 libplacebo 之间插入其他 vf.

P.S. 如果能提供 -loglevel verbose 或以上的完整 log 会更容易使你的问题被解决。注意包含每条 log 前面的 context 信息。

[zshnb]

你好大佬，我用GPU加速的命令./ffmpeg -y -hwaccel cuda -hwaccel_output_format cuda -extra_hw_frames 3 -i UPL_0UOVszENQF.mp4 -vf "hwupload=derive_device=vulkan,libplacebo=format=yuv420p10le:colorspace=bt709:color_primaries=bt709:color_trc=bt709,hwdownload,format=yuv420p" -c:v h264_nvenc -c:a copy UPL_0UOVszENQF.libplacebo.cuda.mp4 同样会报错

Impossible to convert between the formats supported by the filter 'Parsed_hwupload_0' and the filter 'auto_scale_0'
[vf#0:0 @ 0x55f249a12e40] Error reinitializing filters!
Failed to inject frame into filter network: Function not implemented
Error while filtering: Function not implemented
[out#0/mp4 @ 0x55f249a35b80] Nothing was written into output file, because at least one of its streams received no packets.

想请教一下这个可以怎么解决呢

[hym3242]

在创建 filtergraph 时 ffmpeg 会尝试将各个 filter 支持的格式合并，如果不能合并就插入 auto_scale, 你这个应该是两个 filter 支持的格式不同，不要一个用硬件格式一个用软件格式，是不可能合并的，你看 ffmpeg 在 hwupload 和 libplacebo 之间插了一个 auto_scale，很明显是不对的，应该是你的 libplacebo 没有 获得/init 一个硬件 context

[systemyf]

同样问题也出现在，”CUDA硬件加速“这一部分。
没什么好的解决方法，目前没有搜索到的ISSUSE，大部分指向vulkan和cuda不兼容的原因。
比较玄学。

[hym3242]

与其encode再pipe到ffplay，直接mux到sdl2设备是否会更好一些 ( -f sdl2 "")(别忘了在 -i 前面加个 -re 否则解码多快视频就会播放多快)

[yundebian918]

大佬，我按照你写的命令行转码p5的杜比视界结果出先这个提示
[Parsed_libplacebo_1 @ 000002cc915e7a40] [Eval @ 000000cf395fda30] Undefined constant or missing '(' in 'iw'
[Parsed_libplacebo_1 @ 000002cc915e7a40] Unable to parse option value "iw"
Error applying option 'inputs' to filter 'libplacebo': Invalid argument
[vost#0:0/libx265 @ 000002cc93b0dd40] Error initializing a simple filtergraph
Error opening output file outputHLG.mp4.
Error opening output files: Invalid argument
然后我安装提示改了iw和ih之后重新执行代码，结果又出现这个提示
[vf#0:0 @ 000001e171a30dc0] Simple filtergraph 'hwupload,libplacebo=3840:2160:format=yuv420p10le:colorspace=bt2020nc:color_primaries=bt2020:color_trc=arib-std-b67,hwdownload,format=yuv420p10le' was expected to have exactly 1 input and 1 output. However, it had 3840 input(s) and 1 output(s). Please adjust, or use a complex filtergraph (-filter_complex) instead.
[vost#0:0/libx265 @ 000001e16c2674c0] Error initializing a simple filtergraph
Error opening output file outputHLG.mp4.
Error opening output files: Invalid argument
是我哪里搞的不对吗？请大佬明示，感激不尽。

[hym3242]

首先, 不要再叫别人大佬了.
有关你的问题, 如果你不把整个命令行和log po上来我们怎么给你debug呢? 最好在ffmpeg之后紧接着加一个-v verbose这样log会详细许多。你的filtergraph有3840个input这件事就说明ffmpeg在parse你的命令行的时候就出现了问题. 你也可以尝试不缩略filter的param, 而是每个都显式写出,例如libplacebo=w=3840:h=2160:format=yuv420p10le...