Create training-ready transcripts for ML engineers before model runs

Your transcription queue piles unprocessed video rows, delaying dataset updates and blocking model runs. Return cleaned transcripts to the sheet so ML engineers get training-ready data before next run.