《下城见光》AI实验动画 Skill工作流

【生成内容】 # 输出一套可直接进入制作与交付的2.5D厚涂电影风动画项目包，包括已确认完整剧 # 本、内容视觉方向、角色、场景与道具资产清单、角色地基图与三视图、场景母 # 图、正反打画面、独立武器和道具设定图。视频部分按剧本段落提供资产绑定明确 # 的多镜头生成提示词、声音方案、候选视频及每段确认结果。成片阶段可输出选片 # 顺序、背景音乐、主题曲、其他音频、合成母版，以及带台词字幕的最终版本。

【生成内容】 # 启动后选择“上传完整剧本”或“从想法开始共创剧本”。确认剧本与视觉方向后，整理 # 角色、场景和道具清单，依次制作角色地基图、三视图、场景母图及道具设定图。资 # 产确认后，选择视频模型、比例和分辨率，逐段确认声音方式、视频提示词及生成授 # 权。各关键节点均可查看结果，自主确认、修改或重新生成；满意后再进入下一阶 # 段，直至完成选片、配乐、合成与字幕，全程由用户掌握创作决定权。

【生成内容】 # 适用于将一句创意、故事梗概或完整剧本制作成2.5D厚涂电影风动画的创作者与制作 # 团队。可用于剧情短片、品牌故事、概念预告、音乐视觉、IP试播集及社交媒体系列 # 内容。尤其适合需要统一角色、场景与道具外观，连续完成资产图、分段视频、对白 # 音效、选片、配乐、合成和字幕的项目。每个关键节点均可确认、修改或重新生成， # 全程保留创作决定权，让复杂制作更稳定、清晰、完整且可追溯。

Confirmed character reference {{Image 1}} : character concept art sheet, split layout: large close-up bust portrait on the left side, three full-body standing views arranged horizontally on the right side, left to right order: Front view, Profile side view, Back view, full body from head to toe, feet and shoes fully visible, all characters same height, same art style, consistent outfit design, pure white background,

A 32-year-old muscular male human, former security enforcer of the city's control organization turned deserter, disheveled dark brown hair, tired sharp facial features with a small scar near the eye, wearing a complete black-gray tactical combat suit with a high-collar protective jacket, tactical combat trousers, combat boots. His right hand and right forearm are permanently fused/attached with a black hydraulic breach gauntlet featuring three sets of hydraulic pistons, a cold-white energy slot, and a circular empty socket, forming one integrated combat unit that is part of his body design. Neutral standing pose facing forward, full body visible from head to both feet. Reference image {{Image 1}} is only used to help lock in the confirmed facial features, hairstyle, body proportions, basic clothing structure and integrated bodily design of the main subject. No character settings will be added or revised. Do not copy the pose, background, composition, artistic medium, brush strokes, colors and lighting of the reference image.

A tall, heavily built adult male security enforcer belonging to the city’s governing authority, utterly loyal to institutional order. He has short dark hair and stern, rugged facial features, dressed in a full set of dark official combat uniform paired with a tactical vest and robust heavy-duty boots. He stands neutrally facing forward, with the complete full body visible from head to toes. Reference image {{Image 1}} merely assists in finalizing the confirmed facial characteristics, hairstyle, body proportions, basic costume structure and integral bodily design of the subject. No character settings are added or altered; the pose, background, framing, artistic medium, brush texture, color palette and lighting of the reference image must not be replicated.

The interior of a core energy tower beneath the city dome, multi-level annular mechanical architecture rising through the tower's height, an artificial simulated sky dome suspended at the very top emitting cold pale-white false light that conceals the real sky, cold cyan base lighting washing over the tower's structure, red warning light bands sweeping steadily across the mechanical framework, a central control console at the tower's core sealed behind a thick heavy alloy security door that locks the dome's opening mechanism, a deep shaft well descending into the tower's bottom floor, empty environment, no people, no creatures, no silhouettes.

A dense lower mechanical city district beneath the dome, deep vertical shaft between towering structures, crisscrossing thick power cables strung between buildings, corrugated metal rain canopies jutting out at multiple levels, dim flickering signboards mounted on walls, an elevated cargo freight lift platform running along a vertical rail track through the shaft, persistent rain and haze drifting through the narrow gap, empty environment, no people, no creatures, no silhouettes.

An elevated maglev train track and a suspended maintenance platform inside an enclosed dystopian dome city under relentless heavy rain, giant electronic billboards flickering between massive interlocking mechanical tower structures, neon glow diffused into blurred colored blocks by the dense rain haze, volumetric fog, wet reflective metal surfaces, cold cyan machine lighting mixed with red warning lights in the distance, deep industrial architecture receding into the rainy depth, no characters, no people, no silhouettes.

【生成内容】 # 用户可上传参考音频，可以根据已生成视频提取人声，也可依据剧本规划声音；参考只 # 服务当前段落，不跨段误用或覆盖确认内容。生成前展示声音方案与使用位置，经用户 # 确认后生成视频。试听后可保留、修改描述、更换参考或重新生成；未经确认的声音不 # 会进入候选视频与最终成片，确保声画表达准确统一，全程由用户掌握创作决定权。

【生成内容】 # SKILL输出结构化视频提示词， # 视频生成效果稳定。

【生成内容】 # 配乐阶段在分段视频确认及选片后开始，不会提前嵌入音乐，确保对白、环 # 境声和动作音效清晰。系统会分别询问是否需要背景音乐与主题曲，用户可选 # 择不添加，也可协作确定风格、情绪、节奏、乐器、人声及使用位置。试听音 # 乐后，用户均可确认、修改方向、重新生成或替换曲目；未经确认的音乐不会 # 进入成片。最终配乐将根据画面节奏、情绪转折与对白空间完成匹配和混音， # 全程由用户掌握决定权。

【生成内容】 A fatalistic, epic cinematic song with a solo female vocalist (amateur/non-professional raw vocal texture, emotionally weathered rather than polished), telling the story of a deserter breaking free from an eternal rain-locked dome city to reach the true sky. Grand orchestral-electronic fusion: cold synth pads and pulsing bass in the verses building tension, layered with distant industrial percussion like rain on metal; the chorus erupts into soaring strings, driving drums and warm brass, resolving from oppressive minor-key darkness into a triumphant golden major-key release, mirroring light finally breaking through storm clouds. Lyrics (English): \[Verse 1\]\ Seven years the sky was sealed away\ Steel and rain, no dawn, no day\ I was made to guard these walls of night\ Now my hands remember how to fight \[Verse 2\]\ Every order was a chain in disguise\ Every silence was another lie\ I broke the door where the thunder speaks\ Ran until the light found me weak \[Chorus\]\ We were born inside the dark, inside the dark\ Chasing sparks beneath a caged-in sky\ Now the storm gives way, gives way\ Finally the sky belongs to us tonight \[Verse 3\]\ You came to stop me, blade in hand\ Bound by rules you didn't understand\ But every wound we traded in the rain\ Taught us both what freedom truly claims \[Bridge\]\ No more masters, no more chains\ Just two shadows walking out of pain\ Let the tower fall, let the light pour through\ This is the dawn we're breaking into \[Chorus\]\ We were born inside the dark, inside the dark\ Chasing sparks beneath a caged-in sky\ Now the storm gives way, gives way\ Finally the sky belongs to us tonight \[Outro\]\ Finally... the sky belongs to us The vocal delivery should feel raw, slightly imperfect, and deeply human — carrying grief and resolve rather than polished pop perfection. Build from a hushed, tense verse into a wide-open, golden, triumphant chorus.

【生成内容】 # 视频合成在候选片段与播放顺序确认后进行。系统依据选定的视频、配乐和 # 其他音频统一编排，处理段落衔接、声画同步、音量层级并输出母版，不替换 # 镜头、调整剧情或加入未经确认的素材。合成前会展示最终清单取得授权，生 # 成后由用户预览成片，可确认结果、提出修改或重新合成；若某段需要替换， # 只更新相关依赖，不重做已确认内容。未经用户确认的合成版本不会成为最终 # 成片，全程保留决定权。

【生成内容】 # 字幕阶段在成片确认后询问是否添加，不默认生成，也不改动剧情、台词或画面。用 # 户选择需要后，系统依据剧本的说话人、对白、语言和时间位置制作字幕，保持文本 # 准确、时序同步、断句自然，并与画面构图和阅读节奏协调。生成后，用户可逐项检 # 查文字、出现时间、样式与位置，自主确认、修改或重新生成。只有字幕版本得到确 # 认，项目才正式完成；不需要字幕时保留无字幕母版，全程由用户决定。

【生成内容】 # 2.5D厚涂动画｜共创

【生成内容】 # 启动方式： # 加载 Skill 后，输入“开始创作”或发送任意自然语言内容，即可启动工作流。 # 启动后，请选择“上传完整剧本”或“从想法开始共创剧本”。

【生成内容】 # Skill 会根据已确认的剧本提取角色设定，并完成角色图设计。 # 支持根据剧本中的角色设定直接进行文生图。 # 支持上传参考图，结合剧本角色设定生成全新的角色图。 # 也可以将上传的已有成图直接指定为最终角色图，无需重新生成。

【生成内容】 # Skill 会根据已确认的剧本提取场景设定，并完成场景图设计。 # 支持根据剧本中的环境、剧情与氛围设定直接进行文生图。 # 支持上传参考图，结合剧本场景设定生成全新的场景图。 # 也可以将上传的已有成图直接指定为最终场景图，无需重新生成。 # 场景图确认后，可根据镜头需要选择是否生成对应的正反打场景图。

【生成内容】 # Skill 会根据已确认的剧本提取道具与武器设定，并完成对应的资产图设计。 # 支持根据剧本中的外观、材质、结构与功能设定直接进行文生图。 # 支持上传参考图，结合剧本道具或武器设定生成全新的资产图。 # 也可以将上传的已有成图直接指定为最终道具或武器资产图，无需重新生成。 # 资产图确认后，可根据制作需要选择是否生成对应的标准四视图。

【生成内容】 # Skill 会根据已确认的剧本与视觉资产，按剧情段落生成风格统一、叙事连贯的视频。 # 系统会综合角色、场景、道具、镜头语言、人物动作、台词与声音需求，为每段剧情制定完整的视频生成方案。 # 支持选择视频模型、画面比例与分辨率，并在正式生成前查看和确认当前方案。 # 每段视频均可独立预览、确认、修改或重新生成，不影响其他已经确认的创作成果。 # 全部视频确认后，可继续完成选片、配乐、成片合成与字幕制作，全程由用户掌握创作决定权。

【生成内容】 # 开启「自动生成图片/视频」后，Agent 可直接消耗积分并提交生成，无需逐次确认； # 关闭后，每次生成图片或视频前都会等待你的确认。可根据自己的创作习惯进行设置。

【生成内容】 # 建议先让 Agent 根据已确认的剧本、画面节奏与情绪走向，直接输出背景音乐或主题曲 # 提示词。您可以继续要求 Agent 调整曲风、配器、人声、歌词、速度和歌曲结构；确认 # 提示词后，再自行选择平台提供的音乐生成模型进行创作。

【生成内容】 # 平台可用于快速合成；若希望获得更精细的成片控制，建议将已确认的分段视频节点 # 下载后，导入剪映、Premiere Pro 或 DaVinci Resolve 等剪辑软件完成最终合成。 # 这样可以自主调整镜头节奏、转场、声画同步、音量、配乐与字幕。