Week 9: Continuity and production
Combining shots into a sequence, and repairing only the parts of the pipeline that actually failed
26 April 2027
Shots generated separately need to look like they happened in the same room, on the same night. This week is about noticing exactly where they don’t, and fixing only that — not regenerating a sequence that’s already mostly right.
What you’ll be able to do#
- Name what a sequence of separately generated shots must match, and what small variance the eye simply forgives.
- Watch a cut cold and log continuity failures as specific findings, not a vague impression that “something’s off.”
- Decide, for each finding, whether to fix or accept it, and give the reason — then repair only the shot the log actually flagged.
The lecture runs 90 minutes: recap 5, outcomes 5, technique and worked log 45, comparison 15, judgement practice 15, synthesis 5.
What has to match, and what the eye forgives#
Every shot in “Before It Stops” so far was generated on its own, from its own anchor and its own note. Nothing guarantees they agree with each other unless someone checks. Some disagreement matters and some doesn’t:
| Has to match | The eye forgives |
|---|---|
| Wardrobe and prop position between shots of the same moment | Small lighting drift between separately generated takes |
| Geography — where the table, the door, the box stack sit | A slightly different lens feel shot to shot |
| Screen direction — which way she’s facing relative to the phone | Minor skin-tone variance |
The distinction matters because fixing what the eye already forgives wastes a regeneration, the same waste week 6 flagged for scheduling ahead of a pending reference, and week 8 flagged for rewriting a note that wasn’t actually the problem.
Recognising continuity failures precisely#
“It looks a bit off” isn’t a finding — it’s the same shapeless complaint as “more dramatic” in week 7 or “make it better” in week 8, and it’s just as undirectable. A usable finding names the thing: character drift (a prop, a wardrobe detail, a physical feature that changes between shots of the same continuous moment), spatial continuity (geography that contradicts itself — a door on the wrong wall, a box that vanishes without the story compressing time), or temporal continuity (an action that implies a different amount of elapsed time than the cut claims). Naming which of the three a finding is decides what fixing it would actually involve.
Transitions are a directorial choice#
How one shot ends and the next begins is not a software default to accept unexamined. A hard cut claims no time was skipped worth marking. A dissolve claims a passage of time or a connection between two moments. A held frame before the next shot begins claims the moment is still resolving. Choosing between them is the same kind of directorial decision as picking a shot’s tactic in week 8 — it has to be made on purpose, not left to whatever the edit tool inserts by default.
The repair discipline: a real assembly and continuity log#
This actual five-shot assembly of “Before It Stops” runs 31.23 seconds: kitchen wide → caller-screen insert → held reaction close-up → reach and hover → acceptance and listening. The owner supplied the Jimeng footage; the phone insert was generated with Sora 2 and accepted by the owner. The reach-and-hover shot uses Week 7’s Jimeng B candidate for this assembly.
Download the assembly and read the actual continuity log.
- Finding 1 — prop-state continuity: fix. The reach shot originally begins with a dark phone after earlier shots establish a lit incoming call. Trim its first 1.3 seconds, preserving the reach and hover. The new cut-entry frame was checked and shows the phone lit. This repairs the named mismatch without regenerating the performance.
- Finding 2 — lighting variance: accept for this assembly. The reaction close-up is warmer than the reach and acceptance shots. Identity, sweater and eyeline remain broadly consistent; retain the performance for this rough assembly rather than regenerate for the lighting shift alone.
- Finding 3 — screen interface: accept for this assembly, record for the final film. The insert has a dark incoming-call interface while the wider shots show a bright blue-white screen. The call remains active and the insert identifies Theo, so the edit retains the supplied footage. Matching the screen design remains a possible final-film repair.
Hard cuts keep the sequence in one continuous moment; a final half-second fade ends the scene. The log records the decisions and their limits.
Review limits: findings come from actual sampled frames and cut-entry checks, not an unpaused cold-watch or an independent group review. Original audio is retained, but listening, exact “Hey” delivery and sound continuity have not been verified. These are teaching materials, not proof that an individual learner has completed the workshop or final assessment.
Misconception contrast#
| Reacting to a vague impression | Logging a precise finding |
|---|---|
| “Something feels off about this cut” | “The reach shot starts with a dark phone after the call is already lit” |
| Regenerating several shots to be safe | Regenerating only the shot the log names |
| Every variance treated as a problem | Distinguishing what has to match from what the eye forgives |
| Transitions left as the tool’s default | Transitions chosen for what they claim about time and connection |
Guided judgement: fix or accept?#
Watching a cut cold, you notice the phone’s screen brightness reads slightly differently between the caller-screen insert (shot 2) and the reaction close-up that follows (shot 3), both of the same continuous moment. Decide: fix, or accept — and why.
Compare your diagnosis
Screen brightness on a prop, read across two shots of the same continuous moment, is closer to lighting variance than to a geography or wardrobe break — nothing about the story or the space contradicts itself, and nobody watching uncued is likely to register a shift in how bright a phone screen reads for half a second. This sits inside what the eye forgives: accept it, and reserve the regeneration for a finding that actually changes what the audience understands about the space or the moment.
For the workshop, apply the same test to your own cut: for every finding, ask whether it’s a geography, wardrobe, or temporal contradiction, or a variance small enough to forgive — and let that answer decide fix or accept.
What you take to the workshop#
Bring the shots you’ve generated and directed so far for one sequence of your film.
You leave with a cut sequence and a continuity log naming at least one real finding and a defensible fix-or-accept decision — the raw material week 10’s recut and week 11’s sound pass both build on.
Teaching team
- Marisol Quaye — convenor