Jellyfin问题汇总

一、 Jellyfin硬件加速的问题

首先，Jellyfin是支持使用FFMpeg进行视频编码/解码的硬件加速（HWA）的。FFMpeg和Jellyfin可以通过视频加速API支持多种硬件加速实现，例如：

• VA-API是一个视频加速API，它使用libva与本地驱动程序接口用以提供HWA的支持
• QSV使用VA-API的修改（分支）版本，并将其与libmfx及其专有驱动程序连接

以下是官网提供的操作系统与推荐配置的接口

操作系统	推荐的硬件加速
Linux	QSV、NVENC、AMF、VA-API
Windows	QSV、NVENC、AMF
MacOS	VideoToolbox
Android	MediaCodec、OMX
树莓派	OMX

使用HWA的显卡比较

显卡	支持的硬件加速
nVidia	NVENC、VA-API
AMD	AMF、VA-API
Intel	QSV、VA-API
Apple	VideoToolbox
Other	OMX、MediaCodec

1. 核显在你的Linux系统是否正确的加载？

   我的核显型号为：Intel UHD630，Intel在2017年就启动了Media SDK+相关的Linux图形堆栈/驱动程序开源项目。

   组件	Repo	库名
   libVA API	Libva,libva-utils	Libva*.so
   VAAPI Intel Driver	Media-driver,GMMLib	iHD_drv_video.so，igfxcmrt64.so
   Media SDK Library	MediaSDK	libmfxhw64.so and files under /opt/intel/mediasdk

   下面以Ubuntun为例，做构建说明

   1.1. 安装系统依赖

   # sudo apt-get -y install git libssl-dev dh-autoreconf cmake libgl1-mesa-dev libpciaccess-dev libelf-dev
   # mkdir <path-to-working-directory>
   # export WORKING_DIR=<path-to-working-directory>

   1.2. 固件的检查与更新

   1.3. 重新编译libra和libdrm

   ​ 基于prefix的不同配置，你可能需要更新LD_LIBRARY_PATH以便安装后可以找到库文件， 另外，记得首先移除其他的libdrm和libva版本。

   find /usr -name "libdrm*" | xargs rm -rf
   find /usr -name "libva*" | xargs rm -rf

   重建和安装libdrm:

   cp /opt/intel/mediasdk/opensource/libdrm/*/libdrm-2.4.54.tar.bz2 .
   tar xf libdrm-2.4.54.tar.bz2
   cd libdrm-2.4.54
   ./configure --prefix=/usr/local
   make
   make install

   重建和安装libva:

   cd $WORKING_DIR
   git clone https://github.com/intel/libva.git
   git clone https://github.com/intel/libva-utils.git
   cd $WORKING_DIR/libva
   ./autogen.sh --prefix=/usr --libdir=/usr/lib/x86_64-linux-gnu
   make -j4
   sudo make install
   cd $WORKING_DIR/libva-utils
   ./autogen.sh --prefix=/usr --libdir=/usr/lib/x86_64-linux-gnu
   make -j4
   sudo make install

   注意: 请首先编译libdrm，因为libva依赖于libdrm

   编译安装VAAPI Intel Driver

   cd $WORKING_DIR
   git clone https://github.com/intel/media-driver.git
   git clone https://github.com/intel/gmmlib.git
   cd $WORKING_DIR/gmmlib
   cd $WORKING_DIR/media-driver
   cd $WORKING_DIR
   mkdir -p build
   cd $WORKING_DIR/build
   cmake ../media-driver
   make -j8
   sudo make install
   export LIBVA_DRIVER_NAME=iHD
   export LIBVA_DRIVERS_PATH=/usr/lib/x86_64-linux-gnu/dri
   vainfo

   编译安装MediaSDK

   cd $WORKING_DIR
   git clone https://github.com/Intel-Media-SDK/MediaSDK.git
   cd MediaSDK
   mkdir build && cd build
   cmake ..
   make -j8
   sudo make install

   设置必要的环境变量

   cd $WORKING_DIR
   cd MediaSDK/build/__bin/release
   export MFX_HOME=/opt/intel/mediasdk
   export LIBVA_DRIVERS_PATH=/usr/lib/x86_64-linux-gnu/dri
   export LIBVA_DRIVER_NAME=iHD
   ./sample_multi_transcode -i::h264 input.264 -o::h264 out.264

2. 怎么验证做了正确的转码设置？

   要验证使用了正确的库，可以在控制台–>日志，或者终端下查看 /var/log/jellyfin

   grep -A2 'Stream mapping:' /var/log/jellyfin/ffmpeg-transcode-85a68972-7129-474c-9c5d-2d9949021b44.txt

   这返回了以下结果：

   Stream mapping:
   Stream #0:0 -> #0:0 (hevc (native) -> h264 (h264_omx))
   Stream #0:1 -> #0:1 (aac (native) -> mp3 (libmp3lame))

   Stream #0:0用软件解码HEVC（VAAPI Decode也可以说是native），用HWA编码

   Stream mapping:
   Stream #0:0 -> #0:0 (h264 (h264_mmal) -> h264 (h264_omx))
   Stream #0:1 -> #0:1 (flac (native) -> mp3 (libmp3lame))

   Stream #0:0 两者都使用HWA。h264_mmal解码，H264_omx编码

二、Jellyfin字幕方框字的问题

1. Jellyfin本地客户端解码导中文字幕显示方块解决方法

   编辑mpv.conf，放到/Users/用户名/Library/Application Support/Jellyfin Media Player/目录中

   # vim: syntax=config
    
   ###########
   # General #
   ###########
    
   save-position-on-quit
   no-border                               # no window title bar
   msg-module                              # prepend module name to log messages
   msg-color                               # color log messages on terminal
   term-osd-bar                            # display a progress bar on the terminal
   use-filedir-conf                        # look for additional config files in the directory of the opened file
   #pause                                  # no autoplay
   keep-open                               # keep the player open when a file's end is reached
   autofit-larger=100%x95%                 # resize window in case it's larger than W%xH% of the screen
   deinterlace=yes                        # 'auto' does not imply interlacing-detection
   cursor-autohide-fs-only                 # don't autohide the cursor in window mode, only fullscreen
   input-media-keys=no                     # enable/disable OSX media keys
   cursor-autohide=1000                    # autohide the curser after 1s
   screenshot-format=png
   screenshot-png-compression=8
   screenshot-template='~/Desktop/%F (%P) %n'
   hls-bitrate=max                         # use max quality for HLS streams
   ytdl-raw-options=format="[protocol!=http_dash_segments][protocol!=rtmp]"
   #########
   # Cache #
   #########
   cache=yes
   cache-default=81960                     # size in KB  BTW: before is 8000000 it is too big
   cache-backbuffer=25000                  # size in KB
   cache-initial=0                         # start playback when your cache is filled up with x kB
   cache-secs=10                           # how many seconds of audio/video to prefetch if the cache is active
   #############
   # OSD / OSC #
   #############
   osd-level=1                             # enable osd and display --osd-status-msg on interaction
   osd-duration=2500                       # hide the osd after x ms
   osd-status-msg='${time-pos} / ${duration}${?percent-pos:　(${percent-pos}%)}${?vo-drop-frame-count:${!vo-drop-frame-count==0:　Dropped: ${vo-drop-frame-count}}}\n${?chapter:Chapter: ${chapter}}'
   osd-font='Source Sans Pro'
   osd-font-size=32
   osd-color='#CCFFFFFF'                   # ARGB format
   osd-border-color='#DD322640'            # ARGB format
   #osd-shadow-offset=1                    # pixel width for osd text and progress bar
   osd-bar-align-y=0                       # progress bar y alignment (-1 top, 0 centered, 1 bottom)
   osd-border-size=2                       # size for osd text and progress bar
   osd-bar-h=2                             # height of osd bar as a fractional percentage of your screen height
   osd-bar-w=60                            # width of " " "
   #############
   # Subtitles #
   #############
   sub-use-margins
   sub-ass-force-margins
   demuxer-mkv-subtitle-preroll            # try to correctly show embedded subs when seeking
   sub-auto=fuzzy                          # external subs don't have to match the file name exactly to autoload
   sub-paths=ass:srt:sub:subs:subtitles    # search for external subs in the listed subdirectories
   embeddedfonts=yes                       # use embedded fonts for SSA/ASS subs
   sub-fix-timing=no                       # do not try to fix gaps (which might make it worse in some cases)
   sub-ass-force-style=Kerning=yes         # allows you to override style parameters of ASS scripts
    
   sub-font-size=48            # Font Size UPDATE: 2017-02-20
   sub-codepage=utf8:gb18030       # for support Chinese
    
   # the following options only apply to subtitles without own styling (i.e. not ASS but e.g. SRT)
   sub-font="Noto Sans CJK SC Medium"
   #sub-font-size=36
   #sub-color="#FFFFFFFF"
   #sub-border-color="#FF262626"
   #sub-border-size=3.2
   #sub-shadow-offset=1
   #sub-shadow-color="#33000000"
   #sub-spacing=0.5
    
    
   #############
   # Languages #
   #############
    
   #slang=enm,en,eng,de,deu,ger             # automatically select these subtitles (decreasing priority)
   #alang=en,eng,de,deu,ger       # automatically select these audio tracks (decreasing priority)
    
    
   #########
   # Audio #
   #########
    
   audio-file-auto=fuzzy                   # external audio doesn't has to match the file name exactly to autoload
   audio-pitch-correction=yes              # automatically insert scaletempo when playing with higher speed
   volume-max=200                          # maximum volume in %, everything above 100 results in amplification
   volume=100                              # default volume, 100 = unchanged
   ################
   # Video Output #
   ################
   # Active VOs (and some other options) are set conditionally
   # See here for more information: https://github.com/wm4/mpv-scripts/blob/master/auto-profiles.lua
   # The script was slightly modified, functions from scripts/auto-profiles-functions.lua are imported
   # Defaults for all profiles
   vo=opengl
   tscale=linear
   [high-quality]
   profile-desc=cond:is_desktop() and get('width', math.huge) < 3840
   scale=ewa_lanczossharp
   cscale=ewa_lanczossoft
   dscale=mitchell
   scale-antiring=0.7
   cscale-antiring=0.7
   dither-depth=auto
   correct-downscaling=yes
   sigmoid-upscaling=yes
   deband=yes
   [mid-quality]
   profile-desc=cond:(is_laptop() and not on_battery() and get('width', math.huge) < 1920) or (is_desktop() and get('width', math.huge) >= 3840)
   scale=spline36
   cscale=bilinear
   dscale=mitchell
   scale-antiring=1.0
   cscale-antiring=1.0
   dither-depth=auto
   correct-downscaling=yes
   sigmoid-upscaling=yes
   deband=yes
   [low-quality]
   profile-desc=cond:is_laptop() and (on_battery() or get('width', math.huge) >= 1920)
   scale=bilinear
   cscale=bilinear
   dscale=bilinear
   scale-antiring=0.0
   cscale-antiring=0.0
   dither-depth=no
   correct-downscaling=no
   sigmoid-upscaling=no
   deband=no
   [60FPS]
   profile-desc=cond:is_laptop() and get('container-fps', 0) >= 59
   scale=bilinear
   cscale=bilinear
   [4K]
   profile-desc=cond:get('width', -math.huge) >= 3840
   vd-lavc-threads=32
   [4K-inverted]
   profile-desc=cond:get('width', -math.huge) < 3840
   vd-lavc-threads=0
   [default]
   ###################################
   # Protocol Specific Configuration #
   ###################################
   [protocol.https]
   #cache=yes
   #cache-default=500000                    # size in KB
   #cache-backbuffer=250000                 # size in KB
   cache-secs=100                          # how many seconds of audio/video to prefetch
   user-agent='Mozilla/5.0 (Macintosh; Intel Mac OS X 10.11; rv:47.0) Gecko/20100101 Firefox/47.0'
   [protocol.http]
   #cache=yes
   #cache-default=500000                    # size in KB
   #cache-backbuffer=250000                 # size in KB
   cache-secs=100                          # how many seconds of audio/video to prefetch
   user-agent='Mozilla/5.0 (Macintosh; Intel Mac OS X 10.11; rv:47.0) Gecko/20100101 Firefox/47.0'
   [extension.gif]
   cache=no
   no-pause
   loop-file=yes
   [extension.webm]
   #cache=no
   #no-pause
   #loop-file=yes

   需要安装字体：NotoSansCJKsc-Medium.otf.zip