How to Extract FFmpeg Stream by Metadata Tag

This article explains how to isolate and extract a specific audio, video, or subtitle stream from a media file using its metadata tags in FFmpeg. Instead of relying on unpredictable stream index numbers, you will learn how to use FFmpeg’s stream specifiers to target metadata keys like language or title, allowing for precise and automated stream extraction.

The Basic Syntax

To select a stream by its metadata, you use the -map option combined with the metadata specifier m:key:value.

The basic command structure is:

ffmpeg -i input.mkv -map 0:m:key:value -c copy output.mp4

• 0: Specifies the first input file.
• m: Indicates that you are matching by metadata.
• key: The metadata tag name (e.g., language, title).
• value: The specific value of the tag you want to extract.
• -c copy: Copies the stream without re-encoding, ensuring a fast and lossless extraction.

Practical Examples

1. Extract Audio by Language Tag

If you have a video with multiple language tracks and want to extract only the English audio stream, you can target the language tag set to eng:

ffmpeg -i input.mkv -map 0:a:m:language:eng -c copy english_audio.m4a

Note: Adding :a before :m restricts the search specifically to audio streams, preventing the extraction of subtitles that might also share the eng language tag.

2. Extract Subtitles by Language Tag

To extract only the Spanish subtitles from a multi-language MKV file:

ffmpeg -i input.mkv -map 0:s:m:language:spa -c copy spanish_subtitles.srt

Note: :s restricts the selection to subtitle streams only.

3. Extract a Stream by Title Tag

If a stream has a specific title, such as “Director’s Commentary”, you can extract it by targeting the title metadata key:

ffmpeg -i input.mkv -map 0:m:title:"Director's Commentary" -c copy commentary.aac

How to Find Metadata Keys and Values

If you are unsure what metadata tags exist in your file, use the ffprobe command to inspect the streams:

ffprobe -show_streams input.mkv

Look for the TAG: lines in the output to identify the exact keys (like language or title) and values (like eng or Stereo) associated with each stream.