# Cut to Match

> A match-cut planner for short phone videos: list the shots, see every cut read for what carries across it, and get a filming recipe for each cut. https://cut-to-match.skillsafe.ai/

A match cut joins two different moments so something carries across the cut - the same motion, the same shape in the
same place in frame, or the same colour - and the two shots feel like one movement.

## What is free (runs in the browser)

- A shot editor: two to eight shots, each with what is in frame, its shape (round, tall, wide, boxy, a hand, none),
  its place on a 3 by 3 frame grid, its main colour, how it starts and ends (still, left, right, up, down, push in,
  pull back, blur), fast or slow, and its length in seconds.
- A storyboard sheet of pencil panels with a seam chip between every two shots.
- A read of every cut: what is shared (motion, shape, place in frame, colour); the seam it can use (cut the curve,
  zoom-through, inverse zoom-through, rack-focus blur-cut, graphic match, plain cut); and a cut verdict - locked (two
  or more shared, nothing clashing), loose (one shared), broken (nothing shared, motion that doubles back, or a zoom
  that flips from push to pull).
- Sequence checks: a slow side on a motion cut, a motion cut travelling against an earlier one, a pull-back before the
  last cut, more rack-focus cuts than about one per 8 seconds.
- An onion skin of the two frames at each cut (shot A's end in red over shot B's start in cyan) and a velocity strip.
- A verdict by a fixed rule, in order: rethink if any cut is broken; tighten if the page found any gap; ready to shoot
  otherwise.
- Exports: a printable storyboard, Markdown, JSON and a CSV shot list.

## What is metered

One task, `match`, on gpt-terra: one recipe per cut - how shot A should end, how shot B should start, where to make
the cut, and a fix for every loose or broken cut - plus two to four shooting tips and what already works. The page
re-checks every reply (verdict, one entry per cut with the page's seam and verdict, fixes present, numbers and names
only from the input). The recipes use only the shots and notes entered.

## API

See https://cut-to-match.skillsafe.ai/api.html - base https://api.skillsafe.ai/v1/app-api, fields task, idea, aspect,
shots, cuts, facts, figures, notes (JSON-encoded strings), reply contract and worked examples for every verdict.

## Sources

- Built on the SkillSafe skill @heygen-com/cut-the-curve (https://skillsafe.ai/skill/@heygen-com/cut-the-curve): the
  velocity-matched seam catalogue - cut at peak velocity, match direction and speed on both sides of the cut, the
  zoom's scale direction is a sign that must match, one direction of travel, the visible rack-focus cut used sparingly.
- Trend: Wikipedia's "Match cut" article went from 242 daily views on Sep 20, 2026 to a peak of 81,943 on Oct 2 and
  16,091 on Oct 20 (Wikimedia pageviews API).