Guide

Write the target script

Type the sentence you want to hear, hit Submit, and the matcher splices verbatim word and phrase runs from your corpus

The target script is the sentence you want to hear come out of the cut. The matcher reads it, walks every transcribed clip in the corpus, and splices the matched runs together. That's the whole loop.

Type it in

Top-left of the workspace is the script editor. Type. Multi-line is fine. The script saves as you go, so closing the workspace doesn't lose anything.

A few things about how we tokenize:

  • Whitespace splits tokens. Obvious.
  • Dashes, slashes, pipes, brackets, and quotes also split tokens.
  • Apostrophes inside a word stay attached. So "don't" and "it's" are one token each, not two.

Matching is case-insensitive and ignores punctuation when comparing. "Don't worry." in your script matches "dont worry" in a clip's transcript. Your original case and punctuation stay on screen.

The target script editor in the top-left of the workspace with a three-sentence script typed in.

Submit

Click Submit. ReelChest tokenizes the script, indexes the corpus, runs the matcher, and switches the workspace into attribution view. Each matched run gets a tint from a 10-color palette so adjacent clips read at a glance, with the source clip name inline. Words the matcher couldn't find anywhere in the corpus show up neutral gray, flagged as gaps.

The matcher is greedy and verbatim. At every position in your script it tries the longest possible run first, locks it in, and recurses on the rest. Tie-break on equal-length matches: corpus order first, then earliest source word index. Deterministic. Submit twice on the same script and corpus, you get the same cut.

To go back to editing the script, click into it and start typing.

What it doesn't do

  • No fuzzy matching. "I'm gonna" won't match a clip that says "I am going to." Word for word, exactly, or it's a gap.
  • No real-time preview. The matcher doesn't run as you type. Submit is explicit, every time.

If a word is a gap, right-click it. You can substitute a different word into the script (the matcher re-runs), or send the spot to an asset transcript and scan the surrounding context for a word that does exist in your footage.

The honest tradeoff with verbatim-only: it's stricter than guessing-game matching, but you'll never get a splice that says something you didn't film. We picked strict.