# Rule of Six — reference
The reasoning behind each criterion and the doctrine. Source: **Walter Murch**,
*In the Blink of an Eye* — the legendary film editor's (Apocalypse Now, The
Godfather, The Conversation; three Oscars) seminal book on editing, and its
fuller treatment of *why cuts work* beyond the six criteria. Read this for the
*why* behind an edit decision.
## The core claim
The Rule of Six is **not six rules** — it's a **hierarchy of criteria** that
ranks what makes an ideal cut. A perfect cut would satisfy all six at once, but
you rarely can, so the ones at the **top** matter most. Murch assigns each a
rough weight; they sum to 100%:
| # | Criterion | Weight |
|---|---|---|
| 1 | Emotion | 51% |
| 2 | Story | 23% |
| 3 | Rhythm | 10% |
| 4 | Eye-trace | 7% |
| 5 | Two-dimensional plane of screen | 5% |
| 6 | Three-dimensional space of action | 4% |
The startling part: the top two (**emotion + story = 74%**) dwarf the four
"technical" criteria that film schools obsess over. If a cut honors emotion and
story, it's three-quarters right already.
## 6. Three-dimensional space of action — spatial continuity (4%)
Maintaining the **physical reality** of where people and props sit in the space
of a scene. Film-school orthodoxy calls continuity the *most* important thing —
but on a flat screen, gauging 3D space is genuinely hard for the brain to track,
especially across edits. So as long as continuity isn't **egregiously** bad, the
average viewer never notices. That's why it sits at the bottom: prioritize the
higher criteria over strict spatial continuity.
## 5. Two-dimensional plane of screen — photographic grammar (5%)
The grammar of the frame: the **rule of thirds**, **eye-lines**, the **line of
action** (the 180° rule), screen direction. Film school teaches: *never* break
these or your cut becomes incoherent. In reality, movies break them **all the
time** — often deliberately, to provoke unease or a specific emotional response.
The real test is only whether the viewer can still **keep track of the action**
across the shots. Follow the grammar by default; break it *on purpose*, never by
accident.
## 4. Eye-trace — the audience's focus of attention (7%)
Where the viewer's eye is looking, and how it moves, within the frame. **Eye-
trace** guides the eye through an edit by keeping the action / object of interest
in the **same part of the frame** across a cut. This reduces the *friction* of
the cut — it feels coherent and effortless, letting the audience absorb the most
pertinent information fast. It's the hallmark of the **Hollywood montage** and
the **match cut** (yes, including the one you're thinking of).
## 3. Rhythm — the flow of the scene (10%)
The scene's flow, built from lower-order timing elements — **tempo, pacing,
momentum**. Rhythm dictates a scene's energy, elevates its action, and creates an
overall sense of **satisfaction** in a cut. It's best exemplified by cutting to
the **beat** of music, but it lives in everything from dialogue to fight scenes.
Caution: snappy, satisfying edits **must not distract from the story** — rhythm
serves the story, it doesn't upstage it.
## 2. Story (23%)
Not just plot — also **character arcs and themes**. Every cut should **advance
the plot, develop a character, or clarify what the story is about**. Obvious when
said aloud, but films are big, tangled structures with many plot lines, characters
and ideas; identifying a scene's narrative purpose and conveying it concisely is
neither simple nor trivial. Unless you're making a four-hour epic critics find
too long, **let some footage hit the cutting-room floor.**
## 1. Emotion (51%)
The most important, by a wide margin. A technically proficient, well-paced scene
with a coherent story is good — but in a few months the viewer **won't remember
it**. They *will* remember the character's joy, fear, anguish — **how it made
them feel**. So **every cut should convey and evoke emotion**. At the end of the
day, the way films *move* us is what makes them great; the rest is just the
details we need to accomplish that.
## The doctrine: sacrifice from the bottom up
An ideal cut satisfies all six. When they conflict — and they will — **give them
up from the bottom**: 3D space first, then 2D grammar, then eye-trace, then
rhythm. **Never sacrifice emotion; almost never story.** Murch's practical rule:
if a cut nails the emotion but breaks the line of action, take the cut; a
spatially flawless cut that kills the feeling is worthless.
```
keep no matter what ─► Emotion ─ Story ─ Rhythm ─ Eye-trace ─ 2D ─ 3D ─► sacrifice first
```
Murch adds a crucial corollary: **satisfying the higher items obscures problems
with the lower ones — but not the reverse.** If the emotion is right, the story
advances in an interesting way, and the rhythm is right, the audience *tends to be
unaware* of editorial problems in eye-trace, stage line and spatial continuity. A
spatially perfect cut cannot rescue a dead emotion.
## Why cuts work — the blink
*"The cut is a link between two focuses of attention."* Cuts are jumps in space
and/or time, yet audiences follow them effortlessly. Why? Murch's answer is the
**blink**. We experience the world as continuous, but we actually cut it up every
time we close our eyes — and we blink at moments that feel natural: **when a
thought is fully formed.** *"We entertain an idea, and we blink to separate and
punctuate that idea from what follows."*
A film shot presents an idea; the **cut is a blink** that punctuates it. Murch
found he was cutting within a few frames of where **Gene Hackman blinked** in *The
Conversation*, because the blink subconsciously signaled a completed thought. So
the editor is *blinking for the audience* — placing the cut to underscore the
characters' ideas and emotions. Get the rhythm right and *the whole audience
blinks at the same time.*
**Anti-pattern — the "Dragnet" system.** Named after the 50s show that cut after
every line of dialogue. It has a punchy rhythm of its own, but it ignores the
nuance of real conversation (we don't think in metronomic beats). Cut where
thoughts actually complete, not on a mechanical count.
## Movies are dreams — juxtaposition makes meaning
Murch likens editing to **dreaming**: we accept the cut because it resembles how
images are juxtaposed in dreams. In the dark theater we think *"this looks like
reality, but it's so visually discontinuous it must be a dream."* The very
abruptness of the cut produces the dreamlike quality. This is why the **Kuleshov
effect** works — two images placed together create a **third, separate meaning**
that neither shot carries alone. The *order and juxtaposition* of shots is itself
a storytelling tool.
## Don't cut to a too-similar shot — the bees and the 30° rule
Murch's analogy: move a beehive *miles* and the bees reorient fine; move it *two
yards* and they're fatally confused — the environment looks the same, so they
don't re-orient. Editing is the same: **cutting to a shot too similar to the one
before is jarring.** Rule of thumb — the **30° rule**: cut to a camera angle at
least **30° away** from the previous shot (or change the framing/zoom
meaningfully). But note this is a *spatial* concern, and Murch insists spatial
considerations are **not** the editor's primary concern — emotion is.
## The soundtrack rides its own track — split edits and music structure
Murch's six criteria describe the **picture** cut, but two of them — **rhythm**
(#3, "best exemplified by cutting to the beat of music") and, above all, **emotion**
(#1) — are carried more by **sound** than by any single image. So the relationship
between the *soundtrack* and the *picture cut* is a first-class editing decision.
The clearest working-editor's treatment is **Austen Menges**, *How a Pro Video
Editor Uses Music* — a practical companion to Murch:
- **The block-thinking trap.** A timeline welds each clip's video and audio into
one movable block, so beginners **start and stop picture and sound on the same
frame**. Real editing treats video and audio as **independent tracks** that move
on their own and only *occasionally* meet at a sync point — the **split edit**
(J-cut: sound *leads* the picture; L-cut: sound *lingers* past it). It's sleight
of hand: change the picture while the soundtrack holds, or vice-versa; rarely both.
- **Look at the music, don't just listen.** Beyond choosing a track with the right
*feeling*, read its **structure** in the waveform and use the section that fits:
- **Intro** eases the listener in → opens a scene / the hook.
- **Outro** resolves → a **gentle handoff** across a scene change: start the
resolution *offset* from the picture cut so the score carries the viewer over
("picks them up in one scene, sets them down in the next").
- **Variation** — a dynamic quiet↔loud shift — landed **on** the cut pivots the
emotion while the same track keeps playing (lift or settle).
- **The punch is the exception.** A **hard cut** — silence, then the music's first
beat on the first frame — is the one time you *do* sync sound and picture. Punchy,
energetic; reserve it for the moment that earns it.
Why it belongs in the hierarchy: the sound choice *is* the emotion choice (#1), and
overlapping audio across a cut is what makes rhythm (#3) feel inevitable rather than
mechanical — the audio bridge is the aural version of "the whole audience blinks
together." A cut that stops picture, voice and music all at once reads as a machine
(the Dragnet problem, one track deeper).
### Mapping to this repo
`avatar-reel-composer` already gives us a built-in split edit: **one continuous
narration** plays *across* every B-roll / angle cut, so the voice never stops when
the picture does. Exploit it — cut the picture mid-thought over the running voice
rather than resting picture *and* voice together at every sentence pause. The music
**bed** can be a *flat fixed-volume wash* (the "wallpaper music" anti-pattern) **or**
follow a **volume envelope** (`finish_reel.py`) that does editing work. The cut
sheet's job is to say **where** a musical move belongs — an intro beat on the hook
punch, a variation at the emotional pivot, an outro under the close — and those
`sound`/`soundtrack` notes feed the composer directly:
`finish_reel.py --music-from-cutsheet <slug>.cutsheet.json` maps each cut's `sound`
to a move (entrance / lift / settle / duck / resolve) at its scene boundary, or you
author a precise `finish.music_plan` (or use `--music-structure auto` for a tasteful
default). The polish pass's golden flash + whoosh SFX already land *on* B-roll cuts —
that is the "punch" move, so spend it deliberately.
## The editor's job
Murch spends much of the book on the craft's day-to-day, which sharpens how to
plan and *review* an edit:
- **Have a plan.** An editor makes thousands of decisions per scene; when watching
dailies, pick a **single representative still** from each take to make later
take-selection tractable (for reels: the chosen angle crop / key frame).
- **Editing is a dance.** Murch edits *standing up* to internalize the visual
rhythm of what's happening — rhythm is felt in the body, not just seen.
- **See only what's on screen.** He refuses to visit the set or see actors out of
costume, so he judges only *the images the audience will get* — not the intent,
the effort, or what happened during the shoot.
- **Test audiences notice WHEN, not WHY.** If ~80% dislike one scene, the impulse
is to fix or cut *that* scene — but the real problem is often **upstream**: the
audience lacked something they needed to understand it. Fix the missing setup,
not the symptom.
- **Editing starts before post.** The collaboration can begin at the **script, the
storyboard, or the shot list** — which is exactly where this skill's cut sheet
lives. A successful edit is one where *"the shots themselves seem to create each
other."*
## Mapping to this repo
A reel here is a chain of **hard cuts** at scene boundaries (`avatar-reel-composer`).
Each boundary is a cut to run through the hierarchy:
- **Emotion / Story** decide *whether and where* to cut — pin the cut to a reveal,
a punch, a breath, a new idea. This is also why `avatar-reel-composer` places
B-roll to *show* what's said and tightens (emphasis zoom) on the driving lines.
- **Rhythm** = the pacing rules already in `avatar-reel-composer` (hook ~2–3s,
shots ≤ ~6s, alternate angle+zoom on back-to-back talking-heads, cut on the
breath/silence). Its pacing report is the rhythm check.
- **Eye-trace / 2D** = choosing angle crops (`avatar-camera-angles`) and framings
so the face / point of interest stays where the eye expects across a cut, and
eye-lines/screen direction stay consistent (or break on purpose).
- **3D** = spatial continuity; for a talking-head reel it's the least important
and the first to let go.
Companion skills: [`viral-video-script`](../viral-video-script/SKILL.md) writes
the words; [`broll-generator`](../broll-generator/SKILL.md) /
[`broll-finder`](../broll-finder/SKILL.md) supply the Story-beat cutaways;
[`video-scene-analysis`](../video-scene-analysis/SKILL.md) measures an original
reel's rhythm to match.