Long silence at the start throws off Whisper's transcript

Why does a few seconds of dead air at the start garble Whisper's first sentence?

IntermediateRead 6 minDoing: about an hour7 steps

▶ Start

What you'll have at the end

A cleaned transcript with the leading silence trimmed off and no leftover odd symbols in the text.

Before you start

You need

You can already send an audio file to a transcription endpoint and get back its plain-text transcript.

Not covered

Silence or hallucinated fragments turning up later in a long recording, well after the leading dead air is gone; that calls for the wider voice-activity approach in the linked ASR sheet, not a single trim at the front.

Leans on

7 steps

Where to go next

Fix Whisper's misspelled names with a glossary

when only a handful of names or product terms come out spelled wrong and the rest of the transcript is clean throughout, a glossary fixes that; this recipe fixes a bad start caused by dead air, a different problem from a wrong spelling.

Checked 15 Aug 2026

Part of the Multimodal AI cookbook