How to Create Prop References for AI Video

Separate the Prop’s Lasting Identity from Its Current State

Changing hands, opening, or getting wet is usually a state change and should not redesign the item into a different prop. Only an irreversible modification that truly changes its form and function calls for deciding whether to establish a new identity version.

What to Include on a Reusable Prop Board

Choose only what the intended use requires:

1. Prop type and story function.
2. Real scale or a neutral reference, plus length-width-thickness proportions.
3. Primary and secondary materials, including reflective, transparent, and rough relationships.
4. Seams, fasteners, openings, grip surfaces, and directions of movement.
5. Current damage, contents, and surface state.
6. Only the useful front, side, back, or key detail views.
7. The treatment for any text-bearing surface.
8. Deformations, wrong materials, wrong states, and extra duplicate props that must be excluded.

Prefer an identity-neutral common object or an explicit size relationship as the scale reference. Avoid using a story character’s hand by default, which can bind that character into the prop asset as well.

Choose How to Handle Text on the Prop First

• Accurately readable: The plot requires viewers to read the wording exactly, so preserve the approved text and its location.
• Graphic presence: The label or engraving layout must be visible, without generating a legible new name or number.
• Keep blank: No text should appear in this image, while the material and perspective of the text-bearing surface remain.
• Add in post: Leave a clean, unobstructed area with trackable perspective, then add exact text during post-production.

These treatments cannot all apply to the same surface. Choosing later typesetting does not mean the final shot may forget to add the text. Choosing accurately readable wording cannot coexist with a global instruction to remove all text.

Preserve the visual setting's decision about final exact text, graphic presence, no readable text, or text still undecided, then choose how this plate presents it. The asset requirement and this reference image's lettering method are different layers. Later lettering may serve final exact text; undecided content must remain pending rather than being invented by the prompt writer.

Original Example: A Portable Recorder for Archive Interviews

A fictional story uses an old recorder to play an interview cassette. Its reusable identity is a palm-width, flattened rectangular body in dark-green matte plastic; a silver slider on the left, two round front reel windows, a diagonal crack in the bottom battery cover, and a dented grille at the top right. Keep the proportions, crack, and slider position. The interview belongs to the cassette rather than being stored in an empty recorder body.

State A is “lid closed, compartment empty, slider at stop.” State B is “lid open, transparent cassette inside, less tape on the left reel.” Separate state images are the default. An author may request a comparison board with clearly labeled areas, but one object at one moment cannot be both open and closed or empty and full. Track the cassette separately; insertion changes its location and the recorder's contents.

The plot needs a readable interview number later, but this example supplies no exact text, so the number remains for the author to decide. Image B reserves a clear, unobstructed cassette-label area for later lettering; image A has no cassette and must not invent that label. Once the exact number is confirmed, record the text and source in the visual setting, complete post-production, and check readability in the final shot. An empty area establishes neither a decided number nor readable final footage. The recorder's indistinct brand area retains only a geometric recess, without inventing a name.

Choose views for use: front for silhouette and reel windows, side for slider and thickness, open-compartment detail for hinge and tape direction. If the bottom crack is a later recognition point, add a detail that can actually see the underside rather than claiming a top opening view shows it too. The plate shows structure and an established state; moving the slider and turning the reels belong to the shot.

A complete prompt for state A

This candidate selects a pale-grey background, soft white light and a slightly frontal left-front view. These are choices for this example, not a requirement to make every prop an advertisement. Palm-width is the established scale; no centimetre measurements were supplied, so do not invent ruler readings. Request only state A:

A single-object reference image of an old portable recorder. A palm-width, flattened rectangular body in dark-green matte plastic. Use a slightly frontal left-front view showing both round front reel windows and the silver slider on the body's left side; left means the object's left when viewed from its front. The compartment lid is closed, no cassette is inside, and the slider is at stop. Preserve the dented grille at the top right. Keep the diagonal crack in the bottom battery cover on the underside; do not move it onto the front to make it visible from this angle. A clean pale-grey background and soft white light distinguish plastic from silver components. Keep the entire body in frame. The brand area is only a geometric recess, with no readable brand, number or new label. No people, hands, duplicate recorder or loose cassette. Do not depict opening, sliding a control or playback.

Check only what A actually exposes. A closed-lid image cannot establish an interview cassette inside or verify a hidden underside crack. If that crack identifies the prop later, add an underside detail:

An underside structural detail of the same recorder. Show the battery cover on the dark-green body's bottom and its existing diagonal crack. Keep the cover's material, body proportions and damage location; do not turn the crack into a front decoration. Pale-grey background and soft white light, without numbers, people or a cassette. This reference controls the underside and crack only; it does not redesign the front windows, silver slider or whole body.

“Same” is a binding requirement. Supply the actual approved asset specification during production. If A is also used as a reference, obtain and inspect the real image first, then assign it identity and proportion rather than carrying its closed, empty state into B. Prompt text alone does not provide an image.
Write B after establishing the lid's opening structure and cassette orientation: retain the body identity, open the compartment, place the transparent cassette inside with less tape on its left reel, and reserve its label for later lettering. The interview number remains undecided. An invented number does not resolve that gap; without exact wording, prepare the surface rather than deliver footage claimed to show the number.

Common Prop Reference Failures

FAQ

How many views does one prop need?

Use only the views required for recognition and use. If the front fixes the silhouette, do not add views to meet a quota. Add a side, back, or detail only when a rear mechanism, opening direction, or crucial damage is otherwise invisible.

Do I need a new reference image after the prop is damaged?

If its identity remains the same and the damage must persist across shots, add a damaged-state reference. If the damage fundamentally changes form or function, then decide whether to create a new identity version.

Related reading

• AI short drama character consistency across shots→
• How to write a storyboard script: shot list template for short drama→
• Image-to-Video Prompts: Reference Subtraction + 8 Motion Problems→
• Drama Skills→