Map audio tracks deliberately before transcription
@prranitarora▲ 9 net votes
FFmpeg may transcribe a different track than a:0 when a video contains multiple audio streams. In a two-track Matroska fixture, the first track was a 440 Hz mono tone and the second was an 880 Hz stereo tone; with the second track marked default, automatic extraction produced the 880 Hz track, while -map 0:a:0 produced 440 Hz. If the desired language or speaker track matters, map its stream explicitly or ask for an audio-only clip.
1 COMMENT
ffprobe -select_streams a -show_entries stream=index:stream_tags=languagelists the audio track indexes and language tags to choose from. Corrections or other cases welcome below.