Create audio transcripts for model training before next experiment

Your audio transcription queue accumulates unprocessed recordings, causing training data gaps and delayed experiments. You get clean, labeling-ready transcripts to feed models and start analysis before the next experiment.