Communitygithub.com

Akira-TL/akira-video-skills

Design character and scene visual concepts for AI video production; use it when the script only says 'need a person/scene like this' but has not yet formed a stable, generatable visual system for appearance, clothing, materials, colors, space, lighting, and identity.

What is akira-video-skills?

This skill is designed for AI agents and developers working on AI video generation pipelines. It converts vague narrative requirements (e.g., 'a young woman' or 'a futuristic kitchen') into concrete, reusable, and AI-generable visual designs for characters and scenes. It does not generate final images; instead, it produces structured design documents (e.g., `CHR01_design.md`, `LOC01_design.md`) that define stable attributes like silhouette, facial features, costume palette, spatial layout, and lighting direction. The skill works best in conjunction with companion tools like `video-materials` which convert the designs into long-term prompts and reference images (e.g., strict four-view sheets). It emphasizes cross-shot consistency, AI generation stability, and clear separation of high-impact creative decisions (requiring user approval) from routine design completions. Supported platforms include any video-focused AI agent setup (e.g., Claude Code, Cursor); it integrates with a project's existing `VIDEO.md`, `SCRIPT.md`, and reference files in the `references/` directory. The workflow ensures that before any expensive image generation, the visual identity is well-defined, checked for conflicts with existing designs, and approved by the user or system.

Works with~Claude Code~Codex CLI~Cursor
npx skills add https://github.com/Akira-TL/akira-video-skills/tree/HEAD/skills/video/video-design

Ask in your favorite AI

Open a new chat with this agent skill pre-loaded.

Documentation

Video Design

This Skill is responsible for converting narrative and production requirements into reusable, identifiable, and generatable character and scene visual designs. It does not directly handle final image generation, nor does it replace the video-script's character performance / scene narrative definitions.

Design results are permanently stored alongside the corresponding reusable materials, for example:

  • video/materials/characters/CHR01_design.md
  • video/materials/scenes/LOC01_design.md

Subsequently, video-materials will convert the designs into long-term prompts, strict four-view sheets, and reference images.

1. Read Design Constraints First

Read in priority order:

  • VIDEO.md;
  • The current SCRIPT.md / CHARACTERS.md / SCENES.md;
  • The parts of upstream novels, chapters, worldbuilding, or brand materials that are truly relevant to the current object;
  • Existing visual references;
  • Target video style, era, world rules, and generation limitations.

Only inherit already established facts and production decisions. When the script has not defined visual details, this Skill may propose design solutions, but must not retroactively pass off new designs as upstream story facts.

When the project requires a unified overall visual language, first read references/foundation/ART-DIRECTION.md, then proceed to specific character/scene design.

2. Character Design

Character design must not stop at generic descriptions like "young woman", "middle-aged man", "handsome boy".

At minimum, derive from the current narrative function:

  • The character's primary visual identifier in a frame;
  • Overall silhouette;
  • Perceived age, height, and body proportions;
  • Facial structure and stable identifying features;
  • Hair structure;
  • Clothing layers, cut, materials, and accessories;
  • Main colors, accent colors, and contrast relationships;
  • How the character's occupation, era, and life conditions are perceived through appearance;
  • Which features must remain consistent across shots;
  • Which clothing / states are allowed to vary;
  • Which designs are prone to generation drift and should be simplified.

For complex character designs, read references/character/CHARACTER-DESIGN.md as needed.

3. Scene Design

Scene design is more than a noun like "a kitchen" or "a futuristic city"; it must form a space capable of supporting shots and continuity.

At minimum, derive from the current narrative function:

  • What narrative function the scene serves;
  • Spatial scale and overall layout;
  • Architectural / furniture / natural structures;
  • Primary materials;
  • Main and accent colors;
  • Key light source, direction, and time-of-day feel;
  • Critical visual anchor points;
  • How characters and key props can move within the space;
  • Which parts must remain stable across shots;
  • Which details belong only to a specific shot and do not enter the public scene design.

For complex scenes, read references/environment/SCENE-DESIGN.md as needed. When key light direction, day/night, character skin tone, product reflections, or exposure across shots need stability, read references/environment/LIGHTING.md. If the project contains multiple visual worlds, read references/environment/WORLD-DESIGN.md; when main scene spatial relationships or unfamiliar machinery directly affect multiple shots, read references/environment/SPACE-MECHANISM.md.

4. When Design Direction is Uncertain, Create Candidates First

When a choice would significantly change the main character's visual identity, the entire video's visual language, major scenes, or large amounts of downstream material, and upstream materials provide no answer, follow references/foundation/DESIGN-APPROVAL.md to distinguish between "implementation details the Agent can directly complete" and "high-impact creative forks that should first be chosen / explicitly authorized by the user".

When input provides only a function without a clear visual direction, first propose a small number of truly different visual concepts rather than immediately filling in random details.

Candidates must differ genuinely in silhouette, materials, era feel, color, or spatial language, and explain for each:

  • Why it fits the current story / character;
  • Distinctiveness;
  • Multi-shot consistency risk;
  • AI generation stability;
  • Whether it could be easily confused with existing characters / scenes.

Do not generate a large number of candidates when unnecessary. Once the user or current production decision selects a direction, write the chosen design into a long-term design file.

5. Design Files

When formally creating VISUAL_DIRECTION.md, CHRxx_design.md, LOCxx_design.md, or fictional prop designs as needed, use references/foundation/DESIGN-TEMPLATES.md to maintain responsibility boundaries; templates do not require the project to create all files.

Character *_design.md files hold only the currently valid version of the visual design; same for scenes. Git preserves history; do not create pseudo-versions like design_final2.md.

Design files may reference upstream sources and visual references, but must not copy entire paragraphs of fiction, character encyclopedias, or scene histories.

6. Handoff to video-materials

For how to convert character/scene designs into long-term references such as base identity, four-view sheets, clothing/lighting states, follow references/foundation/DESIGN-OUTPUTS.md; these are selected based on real shot requirements, not requiring a full set for every object.

Once a design is confirmed:

  1. video-materials builds long-term image prompts from the design;
  2. When a stable identity/structure is needed, generates reference images following strict four-view templates;
  3. The returned images and prompts are co-located with the design files or placed in the same category directory;
  4. Shot-specific actions, cinematography, and temporary states remain the responsibility of the corresponding shot.

Completion Criteria

A character or scene has been transformed from a "functional description" into a specific, recognizable, cross-shot maintainable, and AI-suitable visual design; the next step can proceed to reference material prompts and four-view generation.

Individual skills in this repo

This repo contains 12 individual skills — each has its own dedicated page.

Akira-TL/akira-video-skills

为 Runway 当前视频模型编译和检查生成提示词;当 video-generation 需要使用 Runway(当前主线 Gen-4.5)的 Text to Video 或 Image to Video,并需要 Runway 专用动作、摄影和时序表达时使用。

Akira-TL/akira-video-skills

为 ByteDance / BytePlus Seedance 当前视频模型编译和检查提示词;当 video-generation 需要使用 Seedance 2.5 的参考素材、时间线、同步声音、视频编辑、延长或多镜头能力时使用。

Akira-TL/akira-video-skills

为 Google Veo 视频生成编译和检查提示词;当 video-generation 需要使用 Veo 的 Text to Video、Image to Video、首尾帧、参考元素或同步声音能力时使用,并先核验具体 Google 使用入口与 model 的当前支持范围。

Akira-TL/akira-video-skills

进入并持续推进一个 AI 视频制作项目;读取或建立 VIDEO.md,根据当前视频脚本、复用素材、镜头、生成结果和整片后期状态,按需路由到 canonical 视频专业 Skill,而不是把所有制作规则塞进一个总提示词。

Akira-TL/akira-video-skills

为品牌短片、产品植入剧情、广告或比赛作品建立产品事实与广告边界;当视频需要准确表达某个具体产品的官方功能、参数、结构、使用方式、品牌素材或产品与剧情关系时使用。

Akira-TL/akira-video-skills

设计并维护 AI 视频中跨镜头复用的角色声音、音色参考、对白声音要求与整片声音关系;当同一角色需要跨镜头保持声音一致,或需要规划语音参考、环境声、音效、音乐与画面的职责边界时使用。

Akira-TL/akira-video-skills

组织整片级剪辑、合成、全片声音和最终成片;当多个已采用镜头需要进入 Premiere Pro、After Effects、DaVinci Resolve 等后期工程,或需要整片字幕、Logo、音乐、混音、比例版本和最终导出时使用。

Akira-TL/akira-video-skills

为 AI 图片或视频生成准备可直接使用的提示词与一次性生成包,并在用户把生成结果带回项目后归档;当需要从正式视频脚本、复用素材和 SHOT.md 编译生成内容、选择参考素材、约定返回文件名或清理临时包时使用。

Akira-TL/akira-video-skills

规划、创建、整理和维护多个镜头会重复使用的视频复用素材;当需要人物/场景/道具/产品提示词、参考图、官方素材、身份与状态参考或其他可重复引用内容时使用。

Akira-TL/akira-video-skills

审核 AI 生成的人物/场景等复用素材或镜头生成结果,判断采用、修改提示词、重生成还是留给后期修复;当生成结果返回项目、需要检查身份/结构/连续性/动作/摄影/产品真实性或选择生成结果时使用。

Akira-TL/akira-video-skills

把小说、章节、故事梗概、品牌 brief 或其他上游内容转换成当前视频自己的脚本层;当需要编写或修订 SCRIPT.md、视频人物表演定义、场景定义、对白或镜头总表时使用。

Akira-TL/akira-video-skills

把 SHOTS.md 中的镜头总览展开成可直接制作的单镜头定义;当需要创建或修改 video/shots/<shot-id>/SHOT.md、镜头连续性、摄影、动作、表演、声音或镜头专属生产约束时使用。

Related Skills