Auto-generated transcripts from Otter, Whisper, YouTube, Zoom, and Fireflies are full of noise — timestamps, "uh", "you know", duplicate speaker labels, and sentence fragments that all add tokens without adding meaning. Transcript Cleaner turns those raw exports into readable prose your AI assistant can summarise, query, or rewrite without wasting context.
How to use it
- 1.
Paste the transcript
Copy your raw transcript from Whisper, Otter, YouTube, or any source and paste it in. SRT, VTT, and plain text all work.
- 2.
Pick what to remove
Toggle timestamps, filler words, duplicate speaker labels, and other artefacts. Every change updates the preview live.
- 3.
Copy the clean version
Grab the cleaned text or Markdown output. The token estimate tells you exactly how much you've saved.
Why transcripts need cleaning before AI
A 60-minute meeting transcript can easily run 15,000+ tokens. Stripping timestamps, filler words, and duplicate names typically cuts that by 30–50% — meaning the same conversation now fits in a smaller context window, costs less per API call, and produces tighter summaries because the model isn't drowning in noise.
Best for
- Meeting summaries with ChatGPT or Claude
- Podcast and interview show notes
- YouTube video Q&A and chapter generation
- Sales-call analysis
- Lecture notes and study guides