Make a Function Card for Each Original Scene
Do not begin with “These two locations look similar.” Record four items for every scene:
| Item | Question to answer |
|---|---|
| Entry conditions | What does each character already know, possess, or promise at this point? |
| Core action | Who pursues, verifies, refuses, exchanges, or conceals? |
| State change | What changes in knowledge, relationships, risk, possession, or authority? |
| Later dependency | Which later scene must use this result? |
A scene devoted only to repeated explanation, travel, or reentering the same issue can be removed or compressed into a transition. Even a brief scene carrying exclusive evidence, a relationship turn, or a means of action cannot simply disappear. Preserve it or transfer its function explicitly into the merged scene.
Atmosphere, enjoying a relationship, and an emotional pause may also be irreplaceable. This card checks causality and trade-offs; it does not remove every scene without new evidence or turn every scene into an argument.
A Scene Merge Must Pass Four Gates
- Time and space connect: Both events can occur in one continuous period and place, without forcing a next-day result into the same minute.
- Agendas collide: The characters’ goals affect one another in the same scene rather than taking turns entering to report separate plots.
- Information order holds: A character obtains a fact before judging or acting on it, and setup still precedes payoff.
- Reaction has room: After an important reveal, refusal, or cost, the affected person can still receive it and change the next step.
If any gate fails, keep separate scenes or merge only part of them. Fewer locations may benefit production, but not at the price of characters becoming suddenly brilliant, props appearing from nowhere, or emotions having no time to turn.
Turn Two Scenes into One Genuine Conflict
After confirming that a merge can work, arrange a new action chain:
- Choose a setting where the relevant people and objects would already gather.
- Let the result of the first agenda trigger the second, rather than relying on “it happened to come up.”
- Put the acquisition of evidence, permission, or tools into the scene.
- Let the opponent counter according to the information that just appeared.
- Exit with a readable state: who knows what, who holds what, and why the next scene occurs.
Merging does not mean attaching two dialogue passages end to end. If Character A delivers information and leaves before Character B enters with another report, there may be one less location, but dramatically these remain two scenes. Make both agendas compete over the same object, time window, decision-making power, or public occasion to create new interaction.
Original Example: Three Matters at the Pre-Contest Instrument Check
The novel and competition rules here are fictional. Luthier Tao Huaiyun brings a student instrument to a charitable competition. Three source scenes serve different functions: finding bridge wood unlike the registered sample at the workshop; a mentor elsewhere suggesting that it resembles missing old stock; and Gu Heng publicly questioning its source after registration. Provenance still needs checking; one score mark establishes neither ownership nor who changed the wood.
The candidate uses the public pre-contest check, with established reasons for the instrument, sample, mentor, and Gu to be there. Tao brings the sample; Gu holds the photograph. Changing locations does not make evidence appear. This example permits voluntary withdrawal. With the relevant holders' agreement, the inspector can retain disputed material and issue a receipt for checking, without deciding ownership at the counter.
- Tao sees two score marks on the bridge's visible side and one on the sample she holds. She pauses submission without knowing the source. The scene need not pretend that tuning exposes a hidden underside.
- The inspector says that verifiable material records are still required by the deadline or the entry must be withdrawn. Changing a line on the form cannot replace that check; Tao must decide whether to continue submitting.
- The mentor says the marks resemble those on his old stock but still require comparison with inventory records. He supplies a search direction without identifying a thief or completing a unique identification.
- Gu demands disqualification and produces a workshop-handoff photograph, arguing that Tao accepted this instrument. It supports that handoff, not who changed material before or afterward. The inspector records it for checking rather than adopting Gu's conclusion.
- Tao voluntarily withdraws this entry and asks for the whole instrument, with its bridge still attached, and the sample to be held for comparison. The mentor agrees to supply inventory records; Gu agrees to include a photo copy; Tao, the current holder of both physical pieces, agrees to their retention. The inspector actually labels them separately, records the bridge/sample discrepancy and the holder of both items, and hands her the receipt. A request alone would not establish completed retention.
The discrepancy, narrowed search, and public challenge remain, each triggering another person's response. The entry is withdrawn and both pieces have traceable retention records, while the person who changed the wood remains unknown. Gu has obtained no admission of responsibility, and the mentor still needs to check records. This is neither permanent disqualification nor a concluded investigation. A later scene checks the handoff and inventory.
The candidate loses the private emotional pause at the workshop, bringing uncertainty and public pressure closer together. If that solitude carries shame, comfort, or a change in trust the author wants to preserve, keep separate scenes or an equivalent pause. Fewer locations do not automatically make the version better.
Common Scene-Compression Failures and Fixes
| Failure | Fix |
|---|---|
| Scenes merge because both take place in an office | Check temporal continuity and whether the agendas affect each other |
| Several people enter in turn to deliver information | Let one result trigger the next person’s countermove or choice |
| A character acts on evidence before obtaining it | Restore the order of acquisition, perception, and understanding |
| Two important reactions collapse into “everyone was shocked” | Preserve the recipient reaction that changes later action |
| Locations decrease, but explanatory dialogue grows longer | Put information into verification, contest, permission, or refusal |
| Props and authority appear from nowhere after the merge | Show who brings them, who controls them, and how they transfer |
| A choice’s consequence disappears to save a scene | Preserve a material loss or obligation in the merged scene’s exit state |
When You Should Keep Two Scenes
Two scenes are stronger apart when time needs to ferment between them, a change of place alters power, a character must decide in solitude, or the latter scene depends on public consequences created by the former.
For example, learning evidence privately and challenging it publicly carry different relationship risks. Forcing them into one public revelation skips the character’s choice about whether to disclose it. In that case, compress travel and repeated explanation while retaining the two functional boundaries. Quality depends on function, not a low scene count; the minimum goal is that every remaining scene be irreplaceable.
FAQ
Does merging scenes always reduce production costs?
It often reduces location changes and repeated establishment, but overloading one scene with characters, props, dialogue, and blocking can also increase difficulty. Make the scene readable first, then compare production benefits.
Can two scenes in the same location be merged automatically?
No. You still need to check time, information order, character agendas, and room for reactions. A shared location solves only a set problem, not the causal structure.
When there are too many scenes, should I cut or merge first?
First decide which subplots and repeated functions will not remain, then merge the surviving scenes. Otherwise, carefully transferred evidence and relationship functions may later disappear with an entire subplot.
Do it with the skill
In Claude Code, use /short-drama-develop to inspect these candidate scenes for merging. For each, list entry conditions, core action, state change, and later dependency, then test continuity of time and space, competing agendas, information order, and room for reactions. Recommend keeping, deleting, or merging each scene without writing script prose. In Codex, use $short-drama-develop.
Read the method: Scene merging, functions, and causal bridges