界面已本地化;技能介绍与说明文档按所选语言由 AI 自动翻译,代码、命令、路径和专业术语保留原文,以原作者文档为准。
探索技能
设计与创意 · SKILL.md

visual-bible

visual-bible

【设计与创意】Define and maintain a film or story project's visual language, moodboard briefs, palette, contrast, lighting logic, texture, composition, camera behavior, location treatment, costume treatment, and recurring motifs.

独立 GitHub 快照 · 自动分类 · 以原作者文档为准

技能说明

GitHub

name: visual-bible description: Define and maintain a film or story project's visual language, moodboard briefs, palette, contrast, lighting logic, texture, composition, camera behavior, location treatment, costume treatment, and recurring motifs. disable-model-invocation: true author: Alan Guice (Badgids) license: Apache-2.0

Visual Bible

Workflow

  1. Read the standalone contract, core contract, brief, canon, world bible, screenplay or story scope, director book if present, and approved visual references.
  2. Write or update 03_preproduction/visual_bible.md.
  3. Define only production-useful rules: aspect intent, palette roles, contrast range, saturation behavior, texture, material emphasis, lighting logic, camera behavior, capture behavior, composition tendencies, environment treatment, costume treatment, recurring motifs, and visual change across the story.
  4. When the story spans visually distinct eras or periods, define an Era / Period Treatments section with palette, light, material, environment, wardrobe, and capture behavior for each applicable period. World facts remain in world.md; this file owns depiction.
  5. Distinguish global rules from scene or sequence exceptions. Describe visible capture behavior instead of inventing camera brands or hardware.
  6. Write 03_preproduction/references/moodboard_briefs.jsonl when moodboard images are useful. Give each frame a job such as palette, material, lighting, composition, architecture, wardrobe, or texture.
  7. Use reference-assets to assign stable reference roles to approved moodboard or style images.
  8. Do not use vague taste words when a visible property can be specified.

Done

A separate artist, director, or generation agent can reproduce the intended visual language without relying on named-film imitation or chat context.

我的收藏