AI Image Prompt Template: Subject and Composition

Decide the picture's job first

One idea can become a character, location, prop, or scene concept. Choose one job for the first image. A character concept emphasizes silhouette and recognition; a location emphasizes entrances, zones, and scale; a prop emphasizes form, function, and material; a scene emphasizes visible relationships among person, place, and object.

This article uses an original paper-only example: a horizontal opening concept for a fictional story. Night plant keeper Cen Yao stands in an old shopping-center rooftop greenhouse, holding a brass seed case that must be returned. No image was generated, tested, or approved.

Fill this manual card and delete irrelevant rows.

Field	Answer in this example
Picture job	Horizontal opening concept showing character, greenhouse, and seed case together
Subject and state	Cen Yao, short curly hair, dark-green apron, holding a closed brass case in both hands
Location relationship	Subject left of center, glass stair door on the right
Key prop	Palm-width, low rectangular brass case with rounded corners and three lid grooves
Required space	Mist-gray glass area in the upper right for a title added later

Translate mood labels into visible facts

“A mysterious woman in an old greenhouse” gives a theme but little visual direction. Ask what identifies the subject, how person/place/prop relate, and which light, surfaces, and colors carry the mood. Each answer should be visible in the result.

Vague idea	Visible repair
Mysterious woman	Night keeper in her twenties, short curly hair, dark-green apron, looking at the case
Old greenhouse	Condensed glass roof, rust-red ribs, pale wooden benches, glass stair door on the right
Tension	Both hands support the unopened case; cold stair light opposes warm light on her profile
Important prop	Palm-width rounded brass case, three lid grooves, handled edges

A few stable details are enough. Hair, apron color, case silhouette, and grooves support recognition. Do not invent an unsettled badge, brand, or wording.

Set composition, light, material, and text treatment

Composition serves the job. For this example: “horizontal medium-wide view; subject left of center at roughly half the image height; glass stair door fully visible on the right; quiet space in the upper right.”

Describe light and material as relationships: a warm lamp reaches her profile, apron, and case edge; cold stair light traces the door frame; the case is aged muted brass with handled edges; glass is condensed and benches are matte wood.

The title is unsettled, so reserve evenly toned glass in the upper right, free of people and crossing leaves, for later lettering. Do not request readable title text. When a sign or package must carry words, confirm exact wording and location first.

Use exclusions that protect this task: open lid, glowing contents, second person, plants blocking the door, and signs or watermarks in the title area.

Copy the completed prompt

Complete paper prompt:

Horizontal opening-scene concept. Cen Yao, the night plant keeper of an old shopping-center rooftop greenhouse, is in her twenties with short curly hair and a dark-green work apron. She stands left of center, holding a closed, low rectangular brass seed case in both hands and looking down at its lid. The case is palm-width with rounded corners and three parallel grooves on the lid; its aged brass is muted, with lightly polished handling edges. Use a horizontal medium-wide view with the subject at about half the image height. Rust-red steel ribs, a condensed glass roof, and pale wooden planting benches establish the rooftop greenhouse; keep the glass stair door fully visible on the right. A warm work lamp reaches her profile, apron, and the upper edge of the case, while cold white stair light traces the door frame. Reserve an evenly toned mist-gray glass area in the upper right for a title added later, with no person, crossing leaves, sign, or watermark in that area. Keep the case closed and unlit inside. Show Cen Yao alone, and keep tall plants from blocking the stair door.

For a new subject, retain the decision order and replace the facts: job → subject and state → prop or location relationship → composition and reserved space → light and material → text treatment → relevant exclusions. Delete parts that do not serve the picture.

Inspect the actual result against its job

When an actual image exists, compare it with the job and card. Fix the highest-impact mismatch first. If the person is correct but plants hide the door, strengthen door visibility and the plant-height boundary. This checklist has not been executed:

Person, greenhouse, and case have a clear hierarchy.
Hair, apron, closed lid, and three grooves are recognizable.
Subject is left of center, door is visible on the right, and title space remains clear.
Warm and cold light explain values on the person, case edge, and door frame.
The case reads as handled aged brass at a believable two-hand scale.
No extra person, glowing contents, door-blocking plant, accidental sign, or watermark appears.
At close view, inspect hand contact, lid closure, door geometry, and bench overlaps.

When several targets miss, repair subject and relationships first, then composition and title space, then secondary surfaces. Retain each actual image with its prompt so revisions can be compared.

Hand it to the right stage in a short-drama project

In a configured short-drama project, short-drama-image-prompts reads accepted character, location, prop, visual-direction, and prompt-language facts and writes copyable text; it does not generate images. Give it the project path, accepted facts, and picture job.

For project generation, transfer the accepted prompt to short-drama-produce, which previews the exact task before creator confirmation of an external call. A manual prompt can also be copied into the creator's chosen tool and inspected against the same brief.

FAQ

Is a longer AI image prompt always better?

Length follows the job. Keep details needed for recognition, spatial relationships, and likely misreadings. Synonymous adjectives, unrelated backstory, and generic defect lists dilute the hierarchy.

Should text inside the image be written directly into the prompt?

For readable text, confirm exact wording, surface, and location, then inspect the result. When wording is unsettled, reserve a clean area and add it later.

Do it with the skill

In the configured short-drama project /ABS/my-drama, use $short-drama-image-prompts to write one [picture job] prompt for the accepted [character, location, or prop]. Use only accepted project asset facts and the project's prompt language. State subject condition, composition and reserved space, location or prop relationships, light and materials, text treatment, and exclusions relevant to this picture. Deliver copyable text and stop at the prompt stage; do not generate an image. If I later request project production, hand the accepted prompt to short-drama-produce, show the exact task preview, and wait for my confirmation before execution.

Read the method: Purpose, writing method, and production boundary for short-drama asset image prompts · General recipe from asset facts to copyable image prompts

About Drama Skills The short-drama-image-prompts skill on GitHub

All guides