Glossary
Punctuation prediction (also called punctuation restoration) is the step that turns raw ASR output — a lowercase stream of words — into properly cased, punctuated, readable text. Older ASR systems output "the meeting starts at three pm"; modern punctuation models give you "The meeting starts at 3 p.m."
Why ASR doesn't punctuate by default
Spoken language doesn't contain punctuation marks. Periods and commas are an artifact of writing. Classical ASR pipelines treated punctuation as a separate task: train an LSTM or transformer model to take a sequence of words and predict where punctuation marks belong. Useful, but a separate model added latency and complexity.
How modern systems handle it
Whisper trains its decoder on punctuated and cased text — so its output already includes punctuation, capitalization, numerals, and basic formatting. No separate model needed. Quality is generally excellent; failures occur on long sentences without natural pauses (Whisper inserts commas at pause boundaries, so monotone speakers get fewer commas).
Where it still falls short
Quoted speech (smart quotes vs straight quotes, attribution like "she said"), em-dashes vs hyphens, currency formatting ($3.50 vs three dollars and fifty cents), date and number formatting (March 5 vs 3/5 vs 3-5), and ALL CAPS for shouting. We post-process Whisper output for currency and some date formatting, but otherwise trust its native output.
In practice
Whisper raw output for an interview: "so I think the the main thing is that we shipped on Friday." Final formatted output: "So I think the — the main thing is that we shipped on Friday." Whisper handles the disfluency repair and adds correct casing automatically.
Related terms
Get clean punctuated transcripts
Get started — freeLast updated: September 20, 2026
Frequently Asked Questions
What does Punctuation prediction mean in transcription?
A post-processing step that adds commas, periods, and capitalization to raw ASR output, which originally produces unpunctuated streams of words.
Why does Punctuation prediction matter when choosing a transcription workflow?
Punctuation prediction affects how you understand transcript quality, timing, compatibility, or the technology behind speech-to-text results.
Where can I apply Punctuation prediction on Transcript.you?
Use the related workflow at /ai-transcription when you want to see how this glossary concept connects to an actual transcription task.
Browse the full transcription glossary or read the complete guide to AI transcription.