Transcript Cleaner
Paste a YouTube transcript, a Zoom or Microsoft Teams transcript, or the contents of a .vtt or .srt file and get readable text back: timestamps stripped, sound cues and filler words removed if you want, and each speaker's captions merged into turns. It runs in your browser — nothing is uploaded — and you choose how much of the original to keep.
Runs entirely in your browser. Your transcript and any file you open are never uploaded.
Separate with commas or line breaks. Whole words only, any capitalization.
Paste a transcript or open a file to clean it.
How to Use the Transcript Cleaner
- Paste or open the transcript. Copy it from YouTube's Show transcript panel, paste a Zoom or Teams transcript, or open a .vtt, .srt or .txt file. The layout is detected for you; override it if the guess is wrong.
- Choose what to keep. Speaker names can stay as Name: text, in bold for Markdown, or go. Timestamps can go, or stay once at the start of each turn.
- Choose how much to clean. Verbatim keeps every word. The next level removes um, uh, er, erm, hmm and mm; the last also removes you know and I mean when commas set them apart. Add your own words, collapse stutters, and drop [Music] and [Applause].
- Check, then copy or download. The word counts show what was removed, and the speaker table shows each person's turns and words. Read the output against the source before you quote it.
Formats It Reads
| Source | What the transcript looks like | How it is read |
|---|---|---|
| YouTube transcript panel | A timestamp such as 1:02 on its own line, then the words | Timestamp lines start a new caption; the lines after it are its text |
| WebVTT (.vtt) from Teams | A cue number or id, a 00:00:05.000 --> 00:00:09.500 line, then <v Name>words</v> | The voice tag gives the speaker; tags and entities are stripped |
| WebVTT (.vtt) from Zoom | The same cue layout with Name: words inside each cue | A Name: prefix is a speaker when most cues carry one or the name recurs |
| SubRip (.srt) | A number, a 00:00:01,000 --> 00:00:03,000 line, then the words | Dash-led lines in one cue are split into separate speakers |
| Zoom text export | [Name] 00:00:05 on one line, then the words | The bracket names the speaker; the time becomes the turn's start |
| Name and time on a line | Name, two or more spaces or a tab, then 0:05 | The name is the speaker; the time starts the caption |
| Plain text | Paragraphs, optionally Interviewer: … | Paragraphs stay apart; recurring Name: prefixes become speakers |
Verbatim or Clean? Choosing How Much to Remove
Transcription for research sits on a spectrum. A verbatim transcript keeps every false start, um and pause; an intelligent or clean verbatim transcript drops the noise and keeps the sense. Which you need depends on your method: conversation and discourse analysis study exactly these features and need them kept, while a thematic analysis of what people said usually works from a clean version. Whatever you choose, state it in your methods section and apply it to every transcript the same way.
The filler options are deliberately narrow. Um, uh, er, erm, hmm and mm are removed as whole words, so summer and uh-huh are untouched. The tool does not remove like, well or so, which carry meaning often enough that deleting them would change what a speaker said; add any you want gone to the custom list. Repeat collapsing is off by default because some repeats are real (that that, had had).
Worked Example
Choose Sample: Teams .vtt: three cues from two speakers, with an um, a stutter (I I started) and an uh. With filler words removed, repeats collapsed, and timestamps kept per turn:
Cleaned
[0:05] Priya Nair: So, how did you first get into nursing?
[0:10] Tom Hale: I started on the wards in 2019, after my degree.
The words go from 21 to 18, there are 3 removals, and Tom Hale's two captions have become one turn. Because captions are joined before they are cleaned, the uh that opened the second caption is removed in the context of its sentence, not left to leave a stray capital letter.
Limits Worth Knowing
- It cleans; it does not correct. Misheard words in an automatic transcript stay as they are. YouTube's auto-generated captions have no punctuation, so sentence-based paragraphs will not split them.
- Rolling captions. YouTube .vtt files repeat each line in the next cue. A line that repeats the previous cue's last line is dropped; check the result if your file is unusual.
- English fillers. The built-in list is English. Add other languages' fillers to the custom list.
- Speakers come from the file. A transcript with no names gets none. For interviews you can add Interviewer: and Participant: labels yourself; the tool will then treat them as speakers.
- Privacy. Transcripts of interviews are personal data. Everything here happens in your browser, but anonymize before you share or code them.
Frequently Asked Questions
- How do I remove timestamps from a YouTube transcript?
- Open the video, choose Show transcript, select and copy the text, and paste it here. Timestamps on their own lines are recognized automatically and removed, and the captions are joined into flowing text. Switch on the paragraph option to break long stretches every few sentences.
- How do I convert a Zoom or Teams transcript to plain text?
- Download the transcript file from the meeting (Zoom's audio transcript or Teams' .vtt download), open it with the file button or paste its contents, and pick a speaker style. The cue numbers, timing lines and tags are removed, and each speaker's consecutive captions become one turn.
- Can I keep who said what but lose the times?
- Yes, and it is the default: speaker names stay as Name: text and timestamps are removed. Choose Keep one at the start of each turn if you want a time to cite or to find the place in the recording.
- What is the difference between verbatim and clean verbatim?
- Verbatim keeps every filler, false start and repetition exactly as spoken. Clean (or intelligent) verbatim removes the noise while keeping the meaning and the speaker's wording. Use verbatim when the way people speak is your data, and clean verbatim when only what they said matters.
- Will it fix errors in automatic captions?
- No. It removes markup, timestamps and the filler you select, and it joins captions into turns. Wrong words, missing punctuation and wrongly attributed speakers stay as they were, so listen to the recording where accuracy matters, especially for quotations.
- Is my transcript sent to a server?
- No. Parsing and cleaning run in your browser, and files you open are read locally. Nothing is uploaded or stored.