Splitting Audio and Incomplete Words

[Acquil]

Context

As per #14 and #7, larger audio files (>4 min) are split into segments depending on the argument min_per_split (minute per split). This is required for faster transcription.

Problem

• There is a high chance of words getting cut-off midway when the audio is being segmented.
• audiosplitter.split_silence(min_per_split) can be used to circumvent this, however it leads to other problems:
  1. All pauses are removed.
  2. The transcripts, while being more accurate, now fail to match the timestamps of the source video/audio.

Audio files should be segmented without modifying the duration or dropping words mid-way.