Why Film Multiple Versions of the Same Line: How Performance States Respond to Player Choices

Multiple versions of the same line are needed because players change the character’s reason for saying it, not to stockpile footage. With identical wording, “I understand” can express trust, dismissal, a threat, or acceptance after a breakdown. Reusing a single neutral performance makes the character forget what the player just did.

Do not write only “sadder” or “a little angrier.” For each version, write what the character wants to get from the other person at that moment, such as forcing an admission, hiding fear, buying time, or preserving dignity. Then add known information, relational distance, physical condition, and the preceding action. These give the actor a basis for creating repeatable differences.

In 《零点回拨》, the partner says, “Give me the phone”: in a high-trust state, the goal is to check the facts together; in a low-trust state, it is to seize control; in a fearful state, it is to stop the protagonist from playing the recording. The line is identical, but the focus of the gaze, distance, gestures, and speaking pace all differ.

Film a state separately only if viewers can perceive it and it affects continuity. Put relationship, knowledge, injuries, and urgency into a matrix, and identify the combinations that actually change the actor’s actions. Do not shoot a version for every small numerical change in the backend; first group them into a few performance levels that can be directed.

Prioritize separate filming for the first reaction after a choice, changes in forms of address, key actions, and transitions into the shared core scene. Informational dialogue in the middle can be reused if its objective is the same. Concentrate the budget where players expect a response.

Each card includes a state ID, the nodes it comes from, what the character knows, what they believe, their relationship with the other character, their current objective, information they must not reveal, physical conditions, references for speaking pace and distance, and entry and exit actions. Actors, directors, script supervisors, and editors use the same document.

State cards prevent acting toward a predetermined outcome. For example, do not write “This is the bad-ending route,” which would give the character advance knowledge of the future; write “She suspects the protagonist has deleted a recording, but has no evidence yet.” Characters can act only on the information they have in the present.

Have the actor perform three versions consecutively, and let the team guess the states without seeing the labels. If they cannot tell them apart, adjust the objective, blocking, or camera setup first, rather than simply increasing the emotional volume. Close shots can convey subtle differences, medium shots can show differences in action, and differences in voice can be substituted during reverse shots or visual obstructions.

The differences must not be so exaggerated that the character seems like different people. Keep the character’s basic habits, accent, and physical logic stable; let the changes occur in trust, control, and pressure.

Across versions, keep the rhythm, positions, and action marks of shared sections as consistent as possible, while giving actors enough freedom in the differing sections. Announce the state ID when slating each take, and have the script supervisor record the specific choice it comes from and compatible exits. When selecting takes, the director should write not only “good performance,” but also “suitable for a low-trust entry.”

Leave extra footage before and after the identical line and record clean sound so versions can be switched in postproduction. If an action is completely different, specify which exit shot it must be paired with; it cannot be freely mixed with other footage.

A common structure uses a short entry wrapper to respond to the player’s choice, after which the character completes a unified action and enters the shared core scene. A high-trust entry might begin with offering water, while a low-trust entry begins with putting documents away; both ultimately sit in the same place to discuss the core information. An exit wrapper at the end then picks up the state again.

The shared core scene must still avoid forms of address and assertions that conflict with any state. If neutralizing the performance leaves the character with no personality, the branches converge too early, and more separate performance should be retained.

The “best take” in editing software is not necessarily the correct take. Link timelines, filenames, and notes to state IDs, and display the node and entry conditions during review. After replacing a version, run regression checks on every path that references it to prevent a high-trust performance from slipping into a betrayal route.

Define a few stable levels for long-term relationships and give them observable behaviors: when guarded, keep a distance and do not volunteer information; when cooperating, exchange facts but keep personal matters private; when trusting, allow vulnerability and willingly take risks. Add the current objective in each scene, instead of playing “trust” as perpetual gentleness. The dictionary keeps the same character consistent across shooting days and different directing units.

Also record the events that trigger state transitions. If a character has just been betrayed but returns to intimacy in the shared core scene, shooting another version may not solve the problem; the script may be missing a transition. Performance versions can carry differences forward, but cannot repair every gap in logic.

Play multiple versions side by side and ask whether viewers can tell what the character wants to do and whether the character remembers the previous choice, rather than which version is “more dramatic.” If shouting is the only way to distinguish them, return to action, distance, gaze, and rhythm to find more specific differences. After choosing a version, record the reason in the state card and apply it to subsequent performances at the same level.

Next step: choose a passage of dialogue reused across multiple routes, and write one action objective and the entry facts for each necessary version. After rehearsal, have people who do not know the labels guess the states blindly; merge or redesign the versions they cannot identify.