# Transcript Cleaner

Paste a raw transcript or captions and clean it into readable prose: strip timestamps, remove filler words, merge speaker turns, and fix punctuation locally.

## Run it

- **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"`
- **Web:** https://gizza.ai/tools/transcript-clean/
- **Agents:** machine-readable descriptor (parameters JSON Schema) at https://gizza.ai/tools/transcript-clean/tool.json

## Inputs

- `input` — Raw transcript _(field)_
- `filler_level` — Filler removal _(field)_
- `extra_fillers` — Extra fillers _(field)_
- `remove_timestamps` — Remove timestamps _(field)_
- `remove_brackets` — Remove bracketed cues _(field)_
- `merge_speakers` — Merge same-speaker turns _(field)_
- `fix_capitalization` — Fix capitalization _(field)_
- `fix_punctuation` — Fix punctuation _(field)_

## Output

- Result (text)

## Query parameters

Open the tool pre-filled and auto-run via URL:

- `input` — Raw transcript
- `filler_level` — Filler removal
- `extra_fillers` — Extra fillers
- `remove_timestamps` — Remove timestamps
- `remove_brackets` — Remove bracketed cues
- `merge_speakers` — Merge same-speaker turns
- `fix_capitalization` — Fix capitalization
- `fix_punctuation` — Fix punctuation

Example: `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=true`

---

## 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`, or `aggressive`. Aggressive also removes discourse
  markers like `you know`, `I mean`, `basically`, `actually`, and `like`, 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`, and `actually` can 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

<details>
<summary>Does this use an AI model to rewrite my transcript?</summary>

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.

</details>

<details>
<summary>What is the difference between standard and aggressive filler removal?</summary>

`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.

</details>

<details>
<summary>Can it clean SRT or VTT captions?</summary>

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.

</details>

<details>
<summary>Why did a bracketed note disappear?</summary>

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.

</details>

## Related tools

- [Context Trimmer](https://gizza.ai/tools/context-trimmer/): Trim text to an approximate token budget for LLM prompts, keeping the head, tail, middle, or both ends. Runs locally in your browser.
- [Compare Two Lists](https://gizza.ai/tools/list-set-diff/): Compare two lists as sets and see which items are only in A, only in B, or shared — with counts. Ignore case, trim, dedupe, and leading zeros. Runs in your browser.
- [Text Diff](https://gizza.ai/tools/text-diff/): Compare two text blocks and highlight added, removed, and changed lines.
- [Word Counter](https://gizza.ai/tools/word-count/): Count the words, characters, and lines in any block of text. Free, instant, runs entirely in your browser — no sign-up required.
- [High-Pass Filter Audio](https://gizza.ai/tools/audio-highpass-filter/): Apply a high-pass (low-cut) filter to audio in your browser — set a cutoff frequency to strip rumble, hum, and handling noise.