Streams and Stream Specifiers

One file, many streams

A media container holds streams: typically one video, one audio, sometimes a subtitle track, occasionally a 'data' stream (timecode, metadata). A Blu-ray rip might have 1 video + 4 audio (different languages) + 8 subtitle streams.

FFmpeg uses stream specifiers to point at specific streams. The grammar is small but you'll hit it constantly.

The grammar

v = video, a = audio, s = subtitle, d = data, t = attachment.
v:0 = the first video stream. a:1 = the second audio stream.
0:v:0 = the first video stream of input #0 (when you have multiple inputs).
-map 0:a:1 = take audio stream #1 from input #0 and put it in the output.
-c:v libx264 = use libx264 for all video streams in the output.
-c:a:0 copy -c:a:1 aac = copy the first audio stream, re-encode the second to AAC.

By default if you don't -map anything, FFmpeg picks the 'best' video and 'best' audio (highest bitrate / preferred language). When you have multi-track files, be explicit — auto-pick is the source of half the 'wrong audio language' bugs in the world.

Code

List streams, then explicitly map·bash

# What streams does this file have?
ffprobe -v error -show_entries stream=index,codec_type,codec_name,channels,channel_layout \
  -of csv=p=0 input.mkv

# Output (example):
# 0,video,h264,,
# 1,audio,ac3,6,5.1
# 2,audio,aac,2,stereo
# 3,subtitle,subrip,,
# 4,subtitle,subrip,,

# Take video, the SECOND audio (stereo AAC), and the FIRST subtitle
ffmpeg -i input.mkv \
  -map 0:v:0 -map 0:a:1 -map 0:s:0 \
  -c copy output.mkv

Per-stream encoding choices·bash

# Video re-encoded H.265, audio kept untouched, subs dropped
ffmpeg -i input.mkv \
  -map 0:v -map 0:a \
  -c:v libx265 -crf 22 \
  -c:a copy \
  output.mkv

Streams and Stream Specifiers

One file, many streams

The grammar

Code

External links

Exercise

Progress

Comments 0