Clean Up a Transcript: Speaker Labels & Timestamps

Enable JavaScript to use the interactive tool. The method and worked example are available below.

Paste your transcript

Works with a transcript you already have, from a meeting tool, a transcription service, or a copied Voice Memos transcript. It does not transcribe audio.

Transcript text

Up to 500,000 characters. The text stays in this browser.

Rename speakers (2 found)

Speaker 1

Speaker 2

Renaming happens before turns are merged, so giving two labels the same name joins them into one turn.

Cleanup options Keep the timestamp on each turn Merge consecutive turns from the same speaker Remove hesitation sounds Leave a blank line between turns

Transcript language

Removes um, uh, erm, mm and hmm, repeated words such as I-I, “you know”, “I mean”, “sort of” and “kind of” before a comma, and “like”, “right” or “okay” between commas. The verb “like” is kept.

Your cleaned transcript

4 turns in, 3 out · 31 words in, 29 out · 2 removed

Speaker labels recognized: a time in brackets before the name.

Speaker 1: I tried the date filter twice. It took three minutes. Then it timed out.

Speaker 2: And what did you do after that?

Speaker 1: I gave up and exported the whole thing.

Speech is never rewritten, only relabeled, rejoined, or stripped of hesitation sounds. Nothing here decides which lines matter, and the original audio is unchanged.

Runs in this browser. Your inputs are not uploaded or saved automatically.

HOW IT WORKS

Same words, fewer things in the way.

Start with the labels you were given

Meeting tools write speakers as Speaker 1, a name and a colon, a chevron, or a name beside a timestamp. All of those are recognized. A colon line that appears only once, such as a note to yourself, stays as text rather than becoming a speaker.

Rename before you merge

Give each label the name you want printed. Renaming runs first, so if two labels turn out to be the same person, their turns merge into one. Reversing that order would leave you editing the same name twice.

Decide what the timestamps are for

Keep them when you will quote the transcript and need to find the moment again. Drop them when someone is going to read it end to end. Removing hesitation sounds is separate, and it never rewrites a sentence.

A WORKED EXAMPLE

A four-turn interview extract

4 turns in, 3 out · 31 words in, 29 out

The first two turns belong to the same speaker, so they join into one. Removing “Um” and the “like” between commas takes the count from 31 words to 29, and the stutter I-I becomes I. No sentence is rewritten, so a quotation still matches the recording. The example is loaded in the tool above; change it and the counts change with it.

Put it into practice

Keep the words, lose the noise.

Guide

Clean up a transcript