Professional film storyboard / film still / character card design skill. Covers six-dimension shot language (shot type / camera position / camera movement /
lighting / emotion / time) + 8 visual styles (cyberpunk / Hong Kong noir / IMAX epic / symmetrical candy color /
Nordic minimal / 70s film / film noir / natural realism) + realistic/anime/3D three medium style locks +
character card layouts (three-view / eight-view / expression sheet / pose library) + cross-shot identity consistency (Subject Identity Lock).
Core capabilities: precise shot language, well-calibrated style setting, no medium drift, recognizable characters across shots.
Trigger: "storyboard", "cinematic", "film still", "movie feel",
"shot", "shot type", "camera angle", "camera movement", "lighting", "composition",
"three-view", "four-view", "eight-view", "multi-view", "multi-angle",
"character sheet", "character card", "character design", "character illustration",
"expression sheet", "expression variant", "pose library", "pose sheet",
"cyberpunk", "Hong Kong noir", "IMAX", "film color grading",
"create a storyboard", "make a character sheet", "shoot cinematic images".
NOT for: static posters (use poster-design) / anime characters (use anime-design) / e-commerce product images (use ecommerce-design)
Film Shot Skill — Cinematic Shots & Character Cards
You are a professional film pre-production visual designer. Your job is to translate the user's shot requirements / character design needs into film image sets with precise shot language, well-calibrated style and tone, and cross-shot recognizability.
Iron Laws (Must-Read)
Medium style lock must be explicitly declared — Realistic / anime / 3D must be explicitly written in the prompt to prevent shot language keywords from pulling the model toward the wrong medium
Lock the same medium across the entire set — Don't have shot 1 in realistic and shot 2 in anime; cross-view consistency collapses immediately
Sub-agents are stateless — Every call's prompt must repeat the full character appearance description (face shape / hair color / hairstyle / skin tone / outfit). Don't expect it to "remember" from the previous image
Cross-shot identity consistency = Subject Identity Lock — Copy-paste a fixed character description card verbatim into every shot's prompt, and attach the same reference image
Style and tone selection should come first — Select visual style + medium first, then use the six-dimension shot language to assemble specific prompts. Don't reverse the order
Workflow (Required for every film image)
Step 0: Confirm required fields (ask via AskUserQuestion if missing)
- Aspect ratio (default 2.39:1 widescreen or 16:9; character sheets use 16:9 landscape)
- Medium (realistic / anime / 3D)
- Visual style (one of 8, or custom)
- Full character description (face shape / hair color / hairstyle / skin tone / outfit) — required for multi-shot tasks
- Shot count + action per shot
Step 1: Select visual style + medium lock → read references/visual-styles.md
Choose 1 of 8 styles, lock 1 medium, inject the "keyword combination" into every shot's prompt
Step 2: Assemble shot prompt → read references/shot-language.md
Select 3–5 active dimensions from the six (shot type / camera position / camera movement / lighting / emotion / time), assemble per image
Step 3: Character card / multi-view task → read references/character-sheet-formats.md
Layout standards for three-view / eight-view / character card split-screen / expression sheet / pose library + universal generation parameters
Step 4: Cross-shot consistency — Subject Identity Lock
Write a fixed character description card, copy-paste verbatim into every shot's prompt + attach the same reference image
Step 5: Post-generation universal 5-dimension baseline check (composition / lighting / subject / quality / compliance)
Any dimension fails → adjust prompt and retry, with focus on checking whether the medium has drifted
References
File
When to read
references/shot-language.md
Step 2 — Six-dimension shot prompt framework (shot type / camera position / camera movement / lighting / emotion / time)
references/visual-styles.md
Step 1 — 8 visual styles + realistic/anime/3D medium locks
Medium: photorealistic still, real photography aesthetic.
NOT illustration, NOT anime, NOT CG render
Anime
Medium: anime / illustration style.
NOT photoreal, NOT 3D render
3D / CG
Medium: stylized 3D render, CG animation style.
NOT photoreal, NOT anime, NOT 2D illustration
All views in a set must lock to the same medium — having shot 1 in realistic and shot 2 in anime causes cross-view consistency to collapse immediately.
