Transcript Cleaner
Turn messy transcripts and subtitle files into clean, readable text — remove timestamps, convert SRT or VTT to TXT, and get Markdown.
Paste a transcript or upload an SRT, VTT, or TXT file. Remove timestamps and subtitle numbers, merge lines into paragraphs, and download clean TXT or Markdown. Runs in your browser; your wording is never changed.
Transcripts rarely arrive in a readable state. A transcript copied from YouTube has a timestamp on every other line. An SRT file is a stack of numbered cues, each with a timing line and two or three words of text. A WebVTT file adds a header, cue settings, and styling tags, and YouTube's automatic captions repeat each line twice as the text scrolls. Dropped into NotebookLM or a document as-is, all of that clutter becomes part of the text. Transcript Cleaner removes only the markup you choose — timestamps, subtitle numbers, caption tags, repeated captions, speaker labels, and sound annotations such as [Music] — and joins the remaining subtitle lines into readable paragraphs. It never rewrites, paraphrases, or corrects your text: the wording and punctuation you put in are the wording and punctuation you get out. Every option is a switch you can turn back on, the original stays visible next to the result, and everything runs locally in your browser, so a private lecture or interview never leaves your device. Paste text or upload a .txt, .srt, or .vtt file, then copy the result or download it as TXT or Markdown.
How it works
- Paste or upload — Paste transcript text, or drop in a .txt, .srt, or .vtt file. No account is needed, and nothing is uploaded — the file is read in your browser.
- Choose what to clean up — Switch options on or off: timestamps, subtitle numbers, paragraph merging, duplicate captions, caption tags, whitespace, speaker labels, and sound annotations. Your wording is never changed.
- Check the preview — Compare the original and cleaned text side by side with word and character counts, then adjust any option and see the result update.
- Copy or download — Copy the cleaned transcript, or download it as a TXT or Markdown file ready for NotebookLM, your notes, or a document.
Why this helps NotebookLM
Subtitle files and copied transcripts are full of timing lines, cue numbers, and repeated captions. When those are uploaded as a source, NotebookLM reads them as part of the text, which wastes words from the 500,000-word-per-source limit and clutters the passages it quotes back to you.
A cleaned transcript reads as ordinary prose: continuous paragraphs, with speaker changes marked where they exist. That is easier for NotebookLM to summarise and cite accurately, and easier for you to check a citation against.
Because all processing happens in your browser, you can prepare confidential interviews, internal meetings, and unpublished lectures without uploading them anywhere before they reach NotebookLM.
What you need to know
How to remove timestamps from a transcript
Paste the transcript into the box (or upload the file) and leave “Remove timestamps” switched on — that is the default. The tool recognises timestamps in three situations: SRT and WebVTT timing lines such as 00:01:02,500 --> 00:01:05,000, timestamps sitting on a line of their own as in a transcript copied from YouTube (0:05 followed by the text), and timestamps at the start of a line such as [00:12] or 00:12 –. In plain text it only treats clock times as timestamps when they are used consistently, in increasing order, throughout the document. A sentence such as “12:30 is when the lab opens” or a line that is only “1984” is left alone. If you want to keep references, switch the option off and each paragraph keeps a [MM:SS] marker from its first line.
Convert SRT to TXT or VTT to TXT
Upload the .srt or .vtt file, or open it in a text editor and paste it. SRT cue numbers and timing lines are removed, WebVTT headers, NOTE blocks, cue identifiers and cue settings are dropped, and the caption text is kept. With “Merge into paragraphs” on, the short subtitle lines are joined into continuous paragraphs, starting a new paragraph at a speaker change or a pause of three seconds or more. Turn merging off if you want one line per subtitle line instead. Then use Download .txt for a plain text file or Download .md for Markdown. Files saved as UTF-8 or UTF-16 are read correctly; older Windows-1252 files are detected and flagged.
Clean a YouTube transcript
On a YouTube video, open the description, choose “Show transcript”, select the transcript text, copy it, and paste it here. (This tool does not fetch transcripts from YouTube links — you provide the text.) Pasted YouTube transcripts alternate between a timestamp line and a caption line; the cleaner removes the timestamps and joins the captions into paragraphs. If you have an automatic-caption .vtt file, the cleaner also removes the word-level timing tags and the repeated line that YouTube's rolling captions carry from one cue to the next. YouTube's automatic captions have no punctuation or paragraph breaks of their own, so the result is readable but unpunctuated — this tool deliberately does not guess at punctuation.
Prepare a transcript for NotebookLM
NotebookLM accepts pasted text as well as .txt and Markdown files, and a clean transcript gives it far better material to cite than a file full of timing lines. A sensible workflow: clean the transcript here, keep speaker labels if people need to be told apart, and use Download .md if you want speaker names in bold and optional timestamps preserved as references. Each source can hold up to 500,000 words, which covers even very long lectures. Add one transcript per source and name the file after the talk so citations are easy to recognise.
What this tool does not do
Transcript Cleaner is a formatting tool, not an editor. It does not remove filler words such as “um” and “you know”, fix grammar or spelling, add punctuation to unpunctuated captions, identify who is speaking, translate, or summarise. Speaker labels are only recognised when they are already in the transcript: WebVTT voice tags, Speaker 1-style labels, >> markers, and names followed by a colon that repeat through the document. A one-off line beginning “Note:” is not treated as a speaker. What it does is remove structure you don't want and leave every word exactly as it was.
Common use cases
- YouTube transcripts for research — Paste a transcript copied from a YouTube video and get readable paragraphs without a timestamp on every other line.
- Lecture and class recordings — Turn subtitle files from recorded lectures into study notes you can read, search, and add to NotebookLM.
- Interviews and meetings — Keep speaker labels, drop the timestamps, and get a transcript that reads like a conversation. Turn timestamps back on when you need to find a quote in the recording.
- Podcast and video captions — Convert SRT or VTT captions to plain text for show notes, articles, or quotations, without retyping.
- Journalism and archiving — Produce a clean text copy of an interview while the original stays untouched and every word is preserved.
What’s preserved
- Your exact wording and punctuation
- Non-English and right-to-left text
- Sentences that contain numbers or clock times
- Speaker labels (unless you remove them)
- Sound annotations like [Music] (unless you remove them)
- Paragraph breaks in plain-text transcripts
What’s removed
- Timestamps and SRT/VTT timing lines
- Subtitle sequence numbers
- WebVTT headers, NOTE blocks, cue settings
- Caption tags such as <i> and {\an8}
- Exact repeated captions between neighbouring cues
- Extra spaces, stray blank lines, invisible characters
Frequently asked questions
How do I remove timestamps from a transcript?
Paste the transcript or upload the file and keep “Remove timestamps” on. It removes SRT and VTT timing lines, timestamps on their own line (as in a pasted YouTube transcript), and timestamps at the start of a line like [00:12]. Sentences that merely contain a time, such as “12:30 is when the lab opens”, are left alone.
How do I convert SRT to TXT?
Upload the .srt file (or paste its contents), leave the defaults on, and choose Download .txt. Cue numbers and timing lines are removed and the subtitle text is kept. Turn off “Merge into paragraphs” if you want one line per subtitle line.
How do I convert VTT to TXT?
Upload the .vtt file and download the result as .txt. The WEBVTT header, NOTE blocks, cue identifiers, cue settings, and styling tags are removed. WebVTT voice tags such as <v Maria> become speaker labels, which you can keep or remove.
Can this get a transcript from a YouTube link?
No. Transcript Cleaner cleans text you provide; it does not fetch transcripts from YouTube or any other site. On a YouTube video, open “Show transcript”, copy the text, and paste it here. If you want to save YouTube videos and their transcripts to NotebookLM directly, that is what the Sourclip extension is for.
Will it change my wording or punctuation?
No. The tool only removes markup (timestamps, numbers, tags, labels, annotations) and re-flows whitespace and line breaks. It never paraphrases, corrects, or reorders text. Each removal is a switch, and the original stays visible beside the result.
Does it remove filler words, fix grammar, or summarise?
No. It is a formatting tool, not an editor or an AI model. It does not remove “um” or “you know”, correct grammar, add punctuation, or write summaries.
Is my transcript uploaded anywhere?
No. All processing happens in your browser. There is no server upload, no AI service, and no account, and the transcript text is never sent to analytics. Anonymous usage events record only things like the detected format and a rough size range.
Which files and formats are supported?
Pasted text, and .txt, .srt, and .vtt files up to 25 MB, in UTF-8 or UTF-16 (older Windows-1252 files are detected and flagged). Standard SRT and WebVTT timestamps, multi-line cues, non-English text, and right-to-left scripts are supported.
Why are my YouTube captions still missing punctuation?
YouTube's automatic captions don't include punctuation or paragraph breaks, and Transcript Cleaner doesn't guess at them because that would mean changing your text. The cleaner joins lines into paragraphs, but the words stay as they were.
Can I keep the timestamps?
Yes. Turn off “Remove timestamps” and each line (or paragraph, when merging) starts with a [MM:SS] marker — or [HH:MM:SS] for transcripts over an hour. The Markdown download shows these in bold.
How do I use the result in NotebookLM?
Copy the cleaned text and add it to a notebook as pasted text, or download the .txt or .md file and upload it as a source. Each source can contain up to 500,000 words.