How to Keep the Same Object Across Two Shots in a Short AI Video
Make this article actionable
Send the article context into Vife Agent and turn it into a plan, checklist, or draft you can keep working on.
Quick answer
If the same object looks like a different object in shot two, the usual cause is not the tool — it is the description. Shot two was written as a fresh idea instead of as a controlled edit of shot one.
The fix is a locked description: one sentence that fixes the object's shape, color, material, and position in frame, plus an explicit list of what must not change (camera position, focal length, light direction, background). Then you change exactly one variable and copy everything else word for word.
Below is a full workflow using a plain white ceramic mug — no logo, no pattern — on a wooden desk against a light gray wall, with soft light from the left. Shot two keeps the mug identical and changes only the wall to dark gray.
Turn the useful parts into next steps
Vife Agent can convert this guide into a prioritized workflow with tasks, risks, and reusable prompts.
Why the object "changes" between shots
Before fixing anything, it helps to know what you are actually fighting. Most drift comes from four places:
- Rewritten descriptions. If you re-describe the mug in your own words for shot two, small word changes ("white mug" becomes "ceramic cup") can shift shape, proportion, or finish.
- Unstated constants. If you never said the camera stays put, the framing may quietly move, and the mug will read as a different size.
- Multiple changes at once. Changing background and light direction and camera angle in one step makes it impossible to tell which change caused the mismatch.
- Ambiguous object words. "Cup" can be tall, wide, tapered, or handled. "Mug" narrows it, but not enough on its own.
None of these are model failures you can prompt your way out of. They are specification gaps. The workflow below closes them.
The four-step workflow
Step 1: Write one sentence that fixes the subject
Describe the object in terms of shape, color, material, and where it sits in the frame. Keep it to one sentence so you can reuse it verbatim.
For the example:
A plain white glazed ceramic mug, cylindrical with a straight-sided body
and a simple loop handle on the right, no logo or pattern, sitting
upright in the lower-center of the frame on a light wood desk.Notice what this sentence does not do. It does not describe the wall, the light, or the camera. Those belong in the constants list, so they can be changed independently later.
Two habits that help here:
- Name the material, not just the color. "Glazed ceramic" reads differently from "matte white plastic."
- Anchor position relative to the frame, not to the room. "Lower-center of the frame" survives a background change; "near the window" does not.
Step 2: Write the constants list explicitly
This is the step most people skip. Write down every element that must stay identical between the two shots:
MUST NOT CHANGE:
- Camera position: eye level, straight on, no tilt
- Framing: mug occupies roughly the same share of the frame
- Focal length: same lens feel, no zoom
- Light direction: soft light from the left, gentle shadow to the right
- Background: flat light gray wallThe background appears here as a constant for shot one. In shot two you will move it out of this list and into the change list — one item, nothing else.
Step 3: Change one variable per version
Shot two is shot one with a single substitution. Copy the entire description and edit one line.
SHOT 2
Subject: A plain white glazed ceramic mug, cylindrical with a straight-sided
body and a simple loop handle on the right, no logo or pattern, sitting
upright in the lower-center of the frame on a light wood desk.
MUST NOT CHANGE:
- Camera position: eye level, straight on, no tilt
- Framing: mug occupies roughly the same share of the frame
- Focal length: same lens feel, no zoom
- Light direction: soft light from the left, gentle shadow to the right
CHANGED THIS VERSION:
- Background: flat dark gray wall (was light gray)If you also want a different camera angle, do that in a third version, not in shot two. One variable per version is what makes the comparison meaningful.
Step 4: Compare the two frames side by side
Put the two frames next to each other at the same display size and look at them in this order:
- Silhouette first. Squint or blur the images. The mug's outline — body width, height, handle shape — should match. Silhouette is the fastest way to catch a shape that drifted.
- Color and finish. Is the white the same white? Is the surface still glossy-glazed rather than matte?
- Position and scale. Does the mug sit at the same height and horizontal position? Is it the same size relative to the frame?
- Light. Does the shadow still fall to the right? Is the highlight still on the left side of the mug?
- Only then, the background. Confirm the wall changed and nothing else did.
If a check fails, fix the description rather than re-rolling and hoping. Re-rolling without changing the text tends to reproduce the same ambiguity.
Review checklist
Use this before you call a two-shot pair finished.
| Check | What "pass" looks like | How to judge it |
|---|---|---|
Shape | Same silhouette, same body proportions, same handle | Subjective — blur the frames and compare outlines |
Color and material | Same white, same glazed finish | Subjective — compare side by side, not from memory |
Position and scale | Mug sits in the same place at the same size | Subjective — overlay or align the two frames |
Light direction | Shadow and highlight on the same sides | Subjective — check shadow direction, not brightness |
Single change | Only the background differs | Objective — count the differences against your change list |
Readability | A viewer understands both shots with no caption | Subjective — show it to someone with no context |
The "single change" row is the only one you can verify mechanically: diff your two descriptions and confirm exactly one line moved from the constants list to the change list.
Common mistakes
- Rewriting instead of copying. Paraphrasing the subject sentence between shots is the single most common cause of drift. Copy and paste, then edit one line.
- Burying the object in a scene description. If the mug is one clause inside a paragraph about the room, it is easy to lose. Give it its own labeled line.
- Changing light and background together. A new wall color often reads as a new light setup. Separate them into different versions so you know what caused what.
- Describing position relative to the room. "On the desk by the window" breaks the moment the background changes. Anchor to the frame.
- Judging from memory. Two shots viewed minutes apart will look more consistent than they are. Always compare side by side.
- Adding a second object to "help" continuity. A saucer, a spoon, or a plant introduces new variables. Keep the scene minimal until the mug itself is stable.
- Assuming a longer prompt is a better prompt. Length is not the goal; a short, unambiguous, reused description is.
A note on tools
This workflow is about the description, not the generator. You can run it in whatever video tool you already use, and the same discipline applies whether you are generating clips, animating stills, or assembling a product sequence. If you want a single place to produce and organize short product-style clips, the AI product video generator is where that work happens in Vife — bring your locked description and constants list with you, and keep the two frames side by side while you review.
FAQ
How many words should the subject sentence be? As few as it takes to remove ambiguity — usually 25 to 45. If you can delete a word and the object still reads the same, delete it.
What if the mug still changes shape between shots? Add a shape-specific detail, such as "straight-sided body, no taper" or "handle is a simple loop, not a squared handle." Vague nouns are the usual culprit.
Can I change the camera angle in shot two? You can, but do it as a separate version. Changing angle and background at once makes it hard to tell which one broke the match.
Do I need to keep the same background color in both shots? No — that is exactly the variable this workflow is designed to change safely. Keep the light direction constant and let the wall color move.
How do I know the change is "only one variable"? Diff your two descriptions. If more than one line differs, you changed more than one thing.
Does a detailed prompt guarantee consistency? No. A locked description reduces ambiguity and makes mismatches easier to diagnose; it does not control how any particular tool renders a scene. Treat it as a specification you can inspect and revise, not a guarantee.
What if the two shots are meant to be different scenes entirely? Then consistency is a design choice, not a requirement. The workflow still helps: decide which properties must carry over (shape, color, material) and which are free to change (background, framing).
Should I write the description before or after choosing the background? Write the subject sentence first, then the constants list, then the change list. Building in that order keeps the object from being defined by its surroundings.