Section 7: Animation and Image to Video · Lesson 4 of 5 (34 of 45) · 60 minutes · Evidence: Motion continuity review
Think about a film scene where you noticed a mistake: a drink that was full and then half empty, a jacket that changed color, a character who walked left in one shot and right in the next. Once you see a slip like that, you stop believing the story. Moving images from generators can make exactly this kind of slip, even inside a single clip.
Production journal: write your answer below.
What must a filmmaker understand before they can maintain face wardrobe hands props background direction and spatial logic over time?
Continuity is the match between shots: the same face, clothes, props, light and setting from one clip to the next. In still images you only had to check one picture at a time. With motion you also have to check time, because things can change during a clip, not just between clips.
Generated video produces temporal artifacts: errors that appear only as time passes. Fingers that multiply, a coat button that disappears, a face that shimmers, a background that slowly reshapes. Identity drift is the most important kind: the character gradually becomes someone else, often by the end of the clip. Drift is easy to miss at normal speed, so you test for it on purpose.
Pause the clip on the first frame, the middle and the last frame. Put the three frames side by side and with the reference image. Compare each anchor in turn. If the glasses differ at the end, you have found identity drift, and you can name the frame where it starts.
Clip A: Mara raises the lamp with her left hand. Clip B, intended to match: her hand lowers the lamp. In the freeze frames of Clip B the lamp is in her right hand and the scarf has changed from mustard to pale yellow. Verdict: two continuity breaks, one in prop handedness and one in wardrobe color. Plan: regenerate with the lamp stated in the left hand and the scarf color restated in the stable prompt, or choose another take.
Creators who make longer sequences from short generated clips spend a surprising share of their time reviewing. Observe how they review, since generating is the fast part and judging is the slow part.
First they replay the clip several times, once at normal speed and then slowly or frame by frame, to look for flicker, bending shapes and changes. They compare a few still frames with the original reference. Then they place the clip next to the one before it and ask whether a stranger would believe the two shots happen in the same place at the same time. When something breaks, they decide whether to regenerate, trim away the bad portion, or replace the shot.
Watch for what the tool cannot do. It has no memory of what Mara wore in the previous clip unless you supply it, and it cannot tell you that a clip has changed meaning. A face that drifts toward looking like a real person, or a gesture that now looks threatening, is a human call.
As you study, record each decision point in your production journal.
| Decision point | What the filmmaker does | What the tool automates | What still needs human review |
|---|---|---|---|
| Replay at normal speed | Watches the clip as an audience would | Plays it back | Whether anything feels off even before examining it |
| Freeze representative frames | Pauses at first, middle and last frames and compares them to the reference | Provides the frames | Identity drift, missing props, shifted colors |
| Compare to neighbouring clips | Places clips side by side and checks light, direction and layout | Nothing; each clip is generated alone | Spatial logic and wardrobe from clip to clip |
| Decide: keep, trim, regenerate or replace | Chooses the cheapest fix that restores continuity | Can regenerate on request | Whether the fix introduced a new problem or changed the meaning |
Studio task: Freeze representative frames and compare identity and object continuity.
Your teacher provides a set of takes of the same short action, some of them with deliberate continuity errors. Work in pairs.
Coaching tips. If you cannot spot a difference, look at the hands, then the edges of the clothing, then the background, because drift often starts there before reaching the face. Use words like “the scarf changes from mustard to pale yellow at about two seconds,” not “it looks weird.” A precise description is what lets you write a fix, and a fix you can write is a fix you can test.
Take name and the clip before it: ______ Reference compared against: ______ Face and hair (holds / shifts / breaks, and at what second): ______ Wardrobe and props (holds / shifts / breaks): ______ Light and background (holds / shifts / breaks): ______ Direction of movement and spatial logic: ______ Meaning check (does motion change what the shot says?): ______ Decision (keep / trim / regenerate / replace) and the fix I will try: ______
Studio task: Review generated takes and plan corrections for continuity failures.
Gather the clips or simulated motion shots you have so far from 7.2 and 7.3. You need at least four. If your access to generation is limited, include shots built from editor moves on stills and review them the same way.
Checklist before you submit:
Identify the exact frame where continuity breaks.
Deliverable: Motion continuity review.
Submit the motion continuity review. The file must use the course naming convention and include enough process evidence to show the student's decisions and revisions.
File name (course naming convention): AIF_S07-L04_motion-continuity-review_v01_YOURINITIALS
| Criterion | Advanced | Proficient | Developing | Beginning |
|---|---|---|---|---|
| The work addresses the stated story or production purpose. | Review links every continuity check to what the scene needs and notes where a break would damage the story. | Review addresses the continuity needs of the scene. | Review lists problems without connecting them to the story. | Review is unrelated to the scene’s purpose. |
| The student applies the required method accurately and produces usable capstone evidence. | Uses the three-frame test on every clip, compares with reference, and classifies each break precisely with its moment. | Applies the test to most clips and identifies breaks with timing. | Reviews some clips or describes breaks vaguely. | No frame comparison is shown. |
| The student identifies at least one quality issue and records a meaningful revision decision. | Plans a specific fix for each break, tests one, and records whether it worked. | Plans fixes and records at least one result. | Fixes are general, with no test. | No fix or revision recorded. |
| The work complies with consent rights attribution privacy disclosure and school policy. | Rejects takes that change identity or meaning misleadingly, avoids drift toward a real person’s likeness, and documents tools and settings. | Rejects clearly misleading takes and documents tools. | Documentation or rejection reasons are partial. | Keeps takes that alter identity or meaning misleadingly. |
Requirement: Reject motion that changes identity or meaning in a misleading way.
Reject any motion that changes who someone is or what a shot means in a misleading way. If a generated face begins to resemble a real person, stop and replace the take. If a gesture turns friendly into threatening, or a character’s skin, age or body changes in a way that stereotypes a group of people, revise or discard it. Keep a note of what you rejected and why, as that record is part of honest disclosure when you say you used AI. Provide a text description of the key action in each clip for viewers who cannot see it.