This repository is no longer maintained.

Use https://github.com/sunrise2575/dist-ffmpeg instead. Thanks.

The automated audio, video, subtitle tool. Currently audio and video is supported.

• Transcode hundreds of video/audio file seamlessly in a computer as the manner of bulk processing

  • Just hit Enter button, and your computer can't be idle before finishing jobs

• Select single stream from multiple streams automatically, transcode/copy each stream, and merge into single file

• Supports user query for general usage conditions (note: query_*.json)

  • movie, drama, anime, music, ... whatever you want to transcode

• golang (for program compile)

• ffmpeg, ffprobe

  • Note: I developed and tested in ffmpeg 3.4.8-0ubuntu0.2 built with gcc 7 (Ubuntu 7.5.0-3ubuntu1~18.04) version

• (Optional) Linux

  • I haven't tested on Windows environment, but if you replace ghw.PCI package to another way, you can utilize my program.

• This program can utilize ffmpeg's NVENC support, so you can use your NVIDIA GPU for transcoding

  • You can see the list of NVENC devices here: https://developer.nvidia.com/video-encode-decode-gpu-support-matrix

    • You can see that older GPUs can't support much codecs.

  • This program doesn't use NVDEC for supporting various codecs without errors, i.e. decode on CPU → encode on GPU)

    • Note: NVDEC supports few codecs like H264, HEVC, VP8, VP9, AV1, ...

• NVIDIA graphics driver (you should see the result of nvidia-smi command)

• NVIDIA NVENC limitation unlock patch: https://github.com/keylase/nvidia-patch

  • That website's Max # of concurrent sessions is a kind of NVIDIA's fraud. It is implemented as GPU driver, so it is a software lock.

  • GPU can have capability to running multiple encoding/decoding session over this software limit. This limit can be unlocked by this patch

  • If you can't/don't patch, you can't run multiple session even you have multiple GPUs

1. Write down the query in *.json file. Your all video/audio files under the selected directory will be processed by this query. The default query_*.json files are included in this package. The query is designed like:

   {
       // "input" section
       "input": {
           // the program scans following files
           "extension": ["mp3", "mkv", "mp4", ...]
       },
   
       // "output" section
       "output": {
           "stream": [
             // "stream" is an array, and the order decides output media stream
             //   For example,
             //     "stream": input query = [{"codec_type":"audio", ...}, {"codec_type":"video", ...}]
             //       -> file's output stream order = ["audio", "video"]
             {
                   // "codec_type" is "video", "audio", and "subtitle"
                   "codec_type": "video",
                   // This example means that:
                   //   The 1st stream will be video
   
                   // You can omit "select_prefer" and "select_priority"
                   // Which means that you choose 0-th video stream whatever it is
   
                   // You can write "copy_if" here
                   "copy_if": {
                       // in this section, you can write regex JSON compare query
                       "codec_name": "^(hevc)$", 
                       "pix_fmt": "^(yuv420p)$"
                   },
   
                   // This stream's example section means that:
                   //   If "codec_name" is "hevc"
                   //      and "pix_fmt" is "yuv420p",
                   //      then the program just copies this stream.
                   //   Otherwise this stream should be transcoded
   
                   // ffmpeg transcoding parameter
                   "ffmpeg_parameter": {
                       "c": "hevc_nvenc",
                       "pix_fmt": "yuv420p",
                       "profile": "main",
                       "level": "auto",
                       "preset": "hq",
                       "qp": "25"
                       // This example section means that:
                       //   ffmpeg ... -c:v hevc_nvenc -pix_fmt:v yuv420p -profile:v main -level:v auto -preset:v hq -qp:v 25 ...
                   },
   
   
                   "temp_file_extension": "mkv" // temp file extension before merging
               },
               {
                   // "codec_type" is "video", "audio", and "subtitle"
                   "codec_type": "audio",
                   // This example means that:
                   //   The 2nd stream will be audio
   
   
                   "select_prefer": {
                       // in this section, you can write regex JSON compare query
                       // regex is combined golang-optional regex and perl regex
   
                       "channels": 2,
                       "tags": {
                           "language": "^(eng|und)$",
                           "title": "(?i)^(?!.*comment).*$"
                           // note: (?i) is golang-style regex, means insensitive
                           //   for additional options like global, multiline, ... you should use golang-style regex
                           // note: ^(?!.*comment).*$ is perl-style regex (i.e. lookahead)
                           //   for core regex, you should use perl-style regex
                       }
                   },
   
                   // "select_priority" is
                   "select_priority": ["tags.title", "tags.language"], // select_prioritty
   
                   // This stream's example section means that:
                   //   If "channels" is 2
                   //     and "language" is "eng" or "und"
                   //     and "title" is not including "comment", the program prefers to choose.
                   //   Especially the program gives most priority for "tags.title",
                   //     second priority for "tags.language"
                   //     and other JSON fields are equally same priority
                   //   Basically, choose only 1 stream for this "codec_type" from entire audio stream candidates,
                   //     therefore the highest ranked stream is selected
   
                   // you can omit "copy_if" section
                   // Which means that that:
                   //   All kinds of audio stream should be transcoded
   
                   // ffmpeg transcoding parameter
                   "ffmpeg_parameter": {
                       "c": "libopus"
                       // What this example section means is that:
                       //   ffmpeg ... -c:a libopus ...
                   },
   
                   "temp_file_extension": "mka" // temp file extension before merging
               }
           ],
   
           // final output extension
           //   If the number of requested stream in query is just one,
           //     and its "temp_file_extension" and this "extension" is same,
           //     then the merging process is skipped, because its behavior is just useless file copying.
           "extension": "mp4"
       }
   }
   

   Note: "codec_type", "select_prefer" and "copy_if" compares its value to the result of following command:

     ffprobe -v quiet -print_format json -show_streams <filename>

2. Run the program by following command, sit back and watch

   (Note: This program recursively found video files under the root folder designated by -root flag.)

   go run . -config <query_JSON_file> -root <your_video_path> -worker <workers>

• This program has been developed for my personal usage. Please be generous for little failure. (But I personally think the program is quite reliable.) You can submit some pull request for fixing bugs and helping me.

• Also, you can fork my repostory, but please notify me when not using this program in personal way (e.g. commercial use, integrate to open source module, ...).

• Subtitle extraction and conversion (SRT, SAMI, VTT, ...)