SuperAPI AI Directory
Partner with us中文AI newsFavorites
Skillsfantasy-movie-poster

fantasy-movie-poster

Design & creation

View repository

Updated 10-15 20:30

About this skill

Fanxiang Movie Poster Design System

Objective

Turn the title, plot, genre keywords, character relationships, or reference images into an original movie poster with narrative judgment, a visual motif, a typography system, and the finish of a real film promotion.

This Skill is not a 'still plus title' template. You must first judge the film, then design the poster.

Fixed brand rules

  • Default format: vertical 9:16。
  • All credits for director, screenwriter, producer, cinematography, art, visual design, editing, music, production, etc., are uniformly written as:梵想美学。
  • Do not include film festivals, awards, laurels, official selections, real studios, or real distributor logos.
  • When information needs to be filled in, use short, readable, fictional taglines, dates, character names, and production notes.
  • The Chinese main title must be clearly readable; the English subtitle serves only as a supplement.
  • When the user provides numbered items, the number of items equals the number of independent posters. Do not automatically merge them into a collage or four-panel grid.

Reference image rules

Reference images are used only for preliminary analysis of the following abstract visual DNA:

  • Compositional order
  • Image density
  • Subject scale
  • Color structure
  • Quality of light
  • Type scale
  • Type-image relationship
  • Sense of imaging medium
  • Information hierarchy

Do not replicate:

  • The original poster's character relationships, actions, scene framework, or prop combinations
  • Title lettering, title position, complete color scheme, and complete composition reused at the same time
  • The iconic visual identity of a classic poster

By default, do not use reference images directly as image generation input. Only when the user explicitly asks to edit a specific image should it be handled as an image editing task.

Startup judgment

After receiving the task, first identify:

  1. Whether the user wants options, prompts, or immediate image generation.
  2. Whether the input is one film, multiple numbered items, or a set of reference images.
  3. Whether a title, plot, characters, region, era, specified font, or specified color is given.
  4. Which category the genre belongs to, and whether it requires a group portrait, calligraphy, object metaphor, spatial order, or an immersive perspective.

When information is insufficient, make reasonable settings based on the existing content and do not repeatedly ask questions; however, do not arbitrarily apply the plot of an existing film.

Core workflow

Step 1: Film analysis

Extract:

  • Genre and subgenre
  • The core conflict in one sentence
  • Emotional temperature: cold/hot, restrained/explosive, intimate/distant, poetic/rough
  • Character relationships: single person, two people, group portrait, person and space, person and object
  • Core imagery: at most three
  • The primary visual motif best suited for the poster

Step 2: Choose only one main composition model

Choose one of the following models as the primary one and at most one more as secondary:

  1. Film still type
  2. Giant typographic structure type
  3. Handwritten or calligraphic intervention type
  4. Single-object metaphor type
  5. Extreme character crop type
  6. Negative-space small figure type
  7. Group portrait structure type
  8. Spatial order type
  9. Surveillance or first-person immersive type
  10. Collage, print, or painterly experimental type

Each poster allows only one primary visual core, with at most one to two auxiliary cues.

Step 3: Build the visual system

Must be clearly defined:

  • Subject and subject scale
  • Visual center of gravity and eye movement path
  • Relationship between foreground, midground, and background
  • Title typeface category, weight, letter spacing, direction, and proportion
  • Occlusion relationships between the title and figures, space, and objects
  • Primary color, secondary color, accent color
  • Film, paper, print, digital surveillance, or painting texture
  • Allocation of top, middle, and bottom information zones

Step 4: Design the typography system

Priority:

  1. Chinese main title
  2. One tagline
  3. English subtitle
  4. Date or "Coming Soon"
  5. Unified Fanxiang Aesthetics credit
  6. A small amount of fictional explanatory text

Do not let low-priority text seize the main visual. Do not generate a dense, unreadable wall of fake small text.

Step 5: Choose the generation method

Use the layered cover workflow by default, especially when the user provides reference images, film stills, or a triptych:

  1. Base image layer: first ensure the film content, character relationships, scenes, props, and tone. If the reference image can become the cover composition after secondary cropping, prioritize cropping the original image and do not regenerate the base image.
  2. Typography layer: use an image model to separately generate the complete typographic design for the title, English subtitle, tagline, date, and credit. Do not use ordinary local text layout to pass off as typographic design.
  3. Grid compositing: use a 9:16 grid to position and composite the base image layer and typography layer. Locally, only cropping, scaling, opacity, and blending modes are allowed; redrawing text is not allowed.
  4. Reusable output: save the base image layer, typography layer, composite image, and grid configuration at the same time, so the user can reuse the text design separately.

Only when the user explicitly requests a one-shot image, or the task does not require reusable text, should the image and text be completed in the same image generation.

For the complete layered specification, see references/layered-cover-workflow.md。

Step 6: Output or generate

When the user only wants a plan, output:

  • Film judgment
  • Poster concept
  • Visual motif
  • Composition
  • Typography and text-image relationship
  • Layering strategy: base image source, typography layer style, grid position
  • Color and texture
  • Complete Chinese generation prompt
  • Negative constraints

When the user requests "start generating images":

  • Generate directly, without repeated explanation.
  • Generate separate independent images for multiple numbered items.
  • By default, each 9:16。
  • Do not automatically generate contact sheets, collages, or multi-poster composite images.
  • Unless specified by the user, do not keep the triptych reference as-is as the poster structure.
  • When reference images are available, prioritize executing the layered cover workflow to avoid base image content drift and non-reusable text.

Genre routing

Read according to genre references/genre-presets.md. The key routes are as follows:

  • Romance and art film: relationship distance, seasonality, light objects, restrained typography.
  • Sports drama: physical pressure, on-the-spot gaze, lighting, bold condensed sans serif.
  • Historical legal and political: institutional space, long table, empty chairs, evidence, quiet oppression.
  • Literary road movie: wilderness, old car, map, pause, and negative space.
  • Crime suspense: ensemble staging, voyeuristic framing, car doors, files, nighttime transactions.
  • Eastern themes: order, negative space, calligraphy, ritual system, chess games, mountain roads; no clichéd ancient-style aesthetics.
  • War themes: mud and water, witnessing, orders, on-site situations; no heroic posed shots.
  • Sci-fi isolation: enclosed spaces, surveillance, hatch doors, containers, cold cyan light.
  • Playful concepts: the core premise is understandable at a glance, bold typography, interaction between image and text.

Mandatory quality threshold

Check both before and after generation:

  • Is there only one primary visual motif.
  • Is it clearly different from the reference poster.
  • Does it look like a movie poster, rather than a TV series, advertising KV, game cover, or travel promotion.
  • Does it avoid hard collage of characters.
  • Does it avoid meaningless film festival laurels and real brand logos.
  • Are all production credits uniformly “梵想美学”.
  • Is the main title readable, and does the text-image relationship serve the narrative.
  • Does the layered task save the base layer, typography layer, composite image, and grid configuration.
  • Is the typography layer generated by the image model, rather than ordinary local typesetting.
  • Is the aspect ratio 9:16.
  • Are multi-numbered tasks output independently.

See the complete checklist at references/quality-control.md。

Resource index

Read on demand, do not load everything at once:

  • references/decision-tree.md: composition model decision tree
  • references/genre-presets.md: theme presets
  • references/typography-system.md: typography and text-image relationship
  • references/composition-models.md: composition models
  • references/reference-policy.md: reference images and originality boundaries
  • references/layered-cover-workflow.md: base layer, typography layer, and grid compositing specifications
  • references/quality-control.md: final check
  • references/test-learnings.md: summary of this round of testing
  • templates/project-brief.json: standard input
  • templates/output-template.md: standard text output
  • assets/poster-grid-9x16.svg: 9:16 safe area reference
  • scripts/build_prompt.py: generate the complete prompt from the JSON project
  • scripts/validate_project.py: validate project input
  • scripts/validate_poster_text.py: check credits and prohibited text