Transcript formatting

Turn raw ASR output into readable transcripts: punctuation, casing, numbers, and paragraphing that stay stable for people and downstream systems.

Formatting makes a transcript useful without changing its underlying meaning. Depending on the selected model and pipeline, it can include punctuation, capitalization, paragraphing, number normalization, and speaker-aware layout.

Keep raw and formatted output distinct

Store the model's raw result when auditability or reprocessing matters. Treat formatted text as a derived representation that can evolve as processors improve.

Design for the consumer

  • Captions benefit from short, quickly finalized segments.
  • Search and analytics benefit from stable normalized text and timestamps.
  • Human-readable transcripts benefit from paragraphs and speaker changes.
  • LLM processing benefits from structured turns rather than one undifferentiated block.

Formatting can be composed with Diarization, Entities, and LLM processing.

On this page