Script Supervision and Continuity: Why Someone Needs to Watch the Details

A viewer may not be able to say exactly what is wrong, but they can sense that "this shot is different from the last one".
That feeling usually comes from continuity errors—props changing position, costumes changing state, the direction of light reversing, or a character's state failing to connect. These problems share one trait: they are hardest to spot during production and easiest to notice while watching.

1. What continuity problems are there?

First, classify them. Continuity problems can be roughly divided into four types.

1. Props and setting

In different shots of the same scene, the position or state of objects is inconsistent.

• In the previous shot, the cup is at the edge of the table; in the next shot, it has moved to the middle
• The number of documents on the table is different
• The weather outside the window changes, from sunny to overcast
• An object appears or disappears in the background

2. Costume and appearance

A character's clothing, makeup, styling, or appearance is inconsistent across consecutive shots.

• In the previous shot, the character is wearing a coat; in the next shot, it is off, with no explanation
• The hairstyle or makeup changes
• The position or state of a wound or stain is different
• An accessory disappears or appears

3. Lighting and time

The lighting conditions do not hold up across consecutive shots.

• The direction of light is opposite in adjacent shots, with light from the left in one and from the right in the other
• The sense of time is inconsistent: even though it is clearly the same scene, some shots feel like daytime while others feel like dusk
• The direction and length of shadows are inconsistent

4. Character state

A character's action, position, or emotional state does not connect.

• In the previous shot, the character is standing; in the next shot, they are suddenly sitting
• Something in their hand disappears
• The emotion jumps from calm to agitated with no process in between
• The eye-line direction does not match, which has already been discussed under axis issues

2. Why it is hard to find but easy to notice

This contradiction is the most troublesome part of continuity problems. The reasons they are hard to find are:

• Production is fragmented. Materials are generated one by one, not shot continuously. During production, attention is focused on whether "this shot itself is right", making it difficult to remember all the details of the previous shots at the same time.
• There is too much information. A project may have dozens or even hundreds of shots, and each shot has multiple details. No one can manage all of it by memory.
• The time span is long. A production cycle may last several weeks, and early details may already have been forgotten.

The reasons they are easy to notice are:

• The viewer's perception is continuous. Viewers watch in order, so differences between earlier and later shots are noticed immediately.
• People are extremely good at recognizing faces. Viewers are unusually sensitive to changes in a character's appearance.
• Subconscious comparison. Viewers do not need to deliberately compare; the brain automatically judges "whether this is the same place and the same person".

So the nature of continuity problems is this: they escape the production team's attention but trigger the viewer's sensitivity.

3. How traditional production solves this

In traditional film and television, this responsibility belongs to the script supervisor. The script supervisor's job is to record the state of every shot: where the props are placed, what the actors are wearing, which direction the light comes from, and what the state of the previous shot was. Every piece of information is written down for later shots to reference. On set, the script supervisor is also responsible for checking before shooting begins—whether the state of this shot connects with the previous shot. The value of this mechanism is that it externalizes "memory" into "records". It does not rely on people remembering; instead, there is an evidence trail that can be checked.

4. Where these responsibilities are in AI production

AI production has no set and no script supervisor. But the need for continuity has not disappeared; only the way to solve it has changed—from "recording" to "asset management". Specifically, several layers correspond to the responsibilities of a traditional script supervisor.

1. Recording character states

Characters may have different states in different scenes: costume changes, injuries, getting wet, or emotional stages.

A traditional script supervisor would record "Scene 15, the protagonist wears a dark coat and has an abrasion on the left arm". In AI production, this corresponds to defining subject states: when extracting a subject, the character's base state and script-related substates are extracted.

Once these states are made explicit, later shots have a basis—when a character at a certain stage is needed, the corresponding state is called, rather than being described again, because describing it again will introduce deviations.

2. Searchability of appearance

States need to be "stored" before they can be reused. NadouPro's character asset library supports filtering by name, tags, gender, age range, skin tone, hair color, body type, and other conditions. With this mechanism, "finding the protagonist's appearance at a certain stage" becomes a filtering action, not a matter of memory or rummaging around. This is similar to the function of a script supervisor's continuity notes in traditional production—both make a state verifiable.

3. A baseline for spatial relationships

In traditional production, consistency in lighting direction and spatial layout is ensured by the script supervisor together with the director of photography.

• In AI production, this part relies on a spatial baseline: panoramic image generation can produce panoramic images used for multi-view scene consistency, setting a "base image" for the space; multi-view generation can change the camera angle and generate multi-view images of the same subject or scene, supporting custom horizontal angles from 0–315 degrees and pitch from -30 to 90 degrees, while also providing preset views such as oblique side, tilted, front low-angle, top, back, and fisheye.

With a set of angle references, checking "whether the lighting direction of this camera position is consistent with the previous shot" becomes a comparison rather than a guess.

4. Shot-level annotation

Continuity checks ultimately have to happen at the shot level. Storyboard nodes carry video storyboard processing and include information such as shot size, movement, and camera angle; field configuration can also adjust parameters such as frame size, depth of field, and lens angle. When each shot's subject reference, state, and angle are written clearly, continuity has a checklist that can be verified.

5. An executable approach

Combining the capabilities above into a workflow, it would roughly look like this:

• Step 1: Create a state list. At the script stage, sort out all state changes for each character, such as costume changes, injuries, and emotional stages, as well as which scenes they appear in.
• Step 2: Turn states into assets. Generate and save the corresponding appearance for each state, and include it in the character asset library.
• Step 3: Establish a spatial baseline. Create multi-angle references for the main scenes first, forming a spatial baseline.
• Step 4: Annotate references in the storyboard. For every storyboard entry, write clearly: which character, which state, which scene, and what angle.
• Step 5: Check one by one before the final cut. Play through in order and compare against the storyboard to check whether props, costumes, lighting, and states connect.

The core of this workflow is to turn "remembering" into "recording", and to turn "going by impression" into "checking against a list". This is also the most essential value of script supervision.

6. Several common misconceptions

• Misconception 1: Thinking continuity problems only matter in large productions. Short videos encounter them too, and because there are fewer shots, the impact of a single mistake is more obvious.
• Misconception 2: Managing by memory. After dozens of shots, memory will definitely fail.
• Misconception 3: Checking only during editing. By the time a problem is found, the material has already been generated, and the cost of revision is high.
• Misconception 4: Not turning states into assets. If a character's state is described again every time, deviations are inevitable.
• Misconception 5: Ignoring the direction of light. This is one of the easiest issues to overlook and one that most affects the viewing experience.

7. A checklist

1. Have all character states been listed? Are any changes missing?
2. Have the states been turned into assets? Can they be called directly instead of described again?
3. Does the scene have a spatial baseline? Can the lighting direction and layout be checked?
4. Has the storyboard been annotated? Does every shot clearly specify the subject, state, and angle?
5. Has everything been checked one by one before the final cut? Has continuity been checked in order?
6. Have the details of props and hands been checked? These small objects are the easiest to get wrong.

Finally

The essence of continuity work is fighting against forgetting. During production, people keep forgetting early details; viewers, however, watch in order, so any inconsistency will be noticed immediately. This asymmetry means continuity cannot be guaranteed by memory. It can only be guaranteed through records and checks. In AI production, this is actually easier to do well than in traditional production, because all materials are digital, searchable, and comparable. The premise is that someone is willing to treat it as a step in the process, rather than leaving it until the final cut to "see if there are any problems".

Contains AI-generated information that may be incorrect. Please verify independently.