Transcript Cleaner
Paste speech-to-text output, SRT/VTT captions, or meeting notes and get a cleaner transcript with fillers, timestamps, cue markers, and repeated speaker turns tidied.
About this tool
Transcript Cleaner turns raw speech-to-text output, SRT/VTT captions, meeting notes, or interview logs into cleaner prose. It strips timestamp scaffolding, removes filler words and hyphenated stutters, drops non-verbal cue markers, merges consecutive turns from the same speaker, and normalizes capitalization and punctuation.
Everything is deterministic and local: no LLM, no network call, and no upload. That makes the result fast and repeatable, but it also means the tool uses fixed rules instead of understanding context.
Worked example
Raw transcript
[00:01:23] Alice: um i think we should ship this
[00:01:26] Alice: uh it looks good [laughter]
Bob: like, agreed
Cleaned transcript
Alice: I think we should ship this. It looks good.
Bob: Like, agreed.
The default standard filler level removes unambiguous vocal pauses (um, uh, erm, hmm),
keeps the potentially meaningful word like, removes timestamps and [laughter], then merges the
two adjacent Alice turns.
Cleanup controls
- Filler removal —
off,standard, oraggressive. Aggressive also removes discourse markers likeyou know,I mean,basically,actually, andlike, which can be useful for a rough draft but may remove meaningful words. - Extra fillers — a comma-separated custom list, applied at any filler level.
- Remove timestamps — drops SRT/VTT cue lines, sequence numbers,
WEBVTT, bracketed clocks, and leading timestamps. - Remove bracketed cues — strips markers such as
[laughter],[inaudible], and(applause). - Merge same-speaker turns — joins consecutive
Name:/[Name]:/>> Name:lines from the same speaker. - Fix capitalization / punctuation — capitalizes sentence starts and
I, spaces punctuation, and adds a final period when a line ends on a word.
Limits
- This is not a grammar rewrite or summarizer. It cleans transcript scaffolding; it does not change sentence order, rewrite awkward phrasing, or infer missing punctuation from audio.
- Aggressive filler removal is opt-in because words such as
like,right, andactuallycan be real content. - Speaker detection is line-based and expects a leading label such as
Alice:or[Alice]:. - Bracket removal is broad: if you need to keep editorial notes in brackets, turn it off.
FAQ
Does this use an AI model to rewrite my transcript?
No. The cleaner is deterministic rule-based code: fixed filler lists, timestamp patterns, speaker label parsing, and punctuation spacing. That keeps it local and repeatable, but it will not rewrite sentences the way a model would.
What is the difference between standard and aggressive filler removal?
standard removes only clear vocal pauses such as um, uh, erm, and hmm. aggressive also
removes discourse markers such as you know, I mean, basically, actually, and like. Use
aggressive for rough cleanup, but review the result because those words can be meaningful.
Can it clean SRT or VTT captions?
Yes. With Remove timestamps on, it drops SRT sequence numbers, cue timing lines containing
-->, the WEBVTT header, bracketed timestamps, and leading clock stamps before cleaning the
remaining caption text.
Why did a bracketed note disappear?
The Remove bracketed cues option strips both square-bracketed and parenthesized spans because
transcripts commonly use them for [laughter], [inaudible], and (applause). Turn that option
off if your transcript uses brackets for content you want to keep.
Developer & Automation Access
Run it from the terminal
Same engine as this page, headless — via the gizza CLI:
gizza tool transcript-clean "[00:01:23] Alice: um i think we should ship this
[00:01:26] Alice: uh it looks good [laughter]
Bob: like, agreed"New to the CLI? Get gizza →
Open it by URL
Pre-fill and auto-run this tool with query parameters — the names match the API/CLI:
https://gizza.ai/tools/transcript-clean/?input=%5B00%3A01%3A23%5D%20Alice%3A%20um%20i%20think%20we%20should%20ship%20this%0A%5B00%3A01%3A26%5D%20Alice%3A%20uh%20it%20looks%20good%20%5Blaughter%5D%0ABob%3A%20like%2C%20agreed&filler_level=standard&extra_fillers=you%20know%2C%20right&remove_timestamps=true&remove_brackets=true&merge_speakers=true&fix_capitalization=true&fix_punctuation=trueMachine-readable descriptor: tool.json — title + parameters JSON Schema for agents.