MayorkingAI 展示了一个完整的 AI 视频制作流程:先用 ChatGPT Image 2 分别生成角色、足球、巴西街头球场和动作编舞分镜,再把这些素材作为参考输入到 Magnific 里的 Seedance 2.0,最终生成一段 10 秒、1080p 的足球花式动作视频。

这个案例的重点不只是最终视频效果,而是它把 AI 视频拆成了“角色一致性、道具一致性、场景一致性、动作时间线”四个可控环节。对于想做稳定 AI 视频的人来说,这比单纯写一个长提示词更有参考价值。

足球花式 AI 视频工作流总览

一、角色设定图:先固定人物身份

作者第一步用 GPT Image 2 生成角色,并把这张图作为参考继续生成 character sheet。这样做的目的是先锁定人物的脸、发型、肤色、体型和服装,减少后续视频生成时人物漂移。

中文整理:创建一个干净、超写实的巴西足球花式玩家角色设定图,保持参考图中的人物身份、脸部、辫发、棕褐色皮肤、瘦而运动型的体格,以及黄色巴西球衣、蓝色短裤、白色中筒袜和花式足球鞋。白色背景,单张图展示正面、侧面、背面全身,以及脸部特写。全身视图使用中性站姿,保持比例一致、真实肤质、清晰细节、柔和棚拍光和专业角色设定图排版。

英文提示词

Create a clean ultra-realistic character sheet of the same Brazilian football freestyler from the reference image. Preserve his identity, face, braided hairstyle, tan skin, slim athletic build, and outfit exactly: yellow short-sleeve football jersey with green trim, blue shorts, white crew socks, and freestyle sneakers. Show him on a pure white background in one single image with four views: full-body front view, full-body side view, full-body back view, and a close-up portrait of his face. Use a neutral standing pose for the full-body views, with consistent proportions, realistic skin texture, sharp details, soft studio lighting, and a professional character sheet layout.

ChatGPT Image 2 生成的角色设定图

二、花式足球:单独生成核心道具

第二步是生成足球道具,而且不是普通比赛用球,而是专门为 freestyle football 设计的花式足球。道具独立生成后,可以在最终视频提示词里作为单独参考,帮助模型保持球的外观一致。

中文整理:一个高端花式足球,略小于常规比赛用球(4 号球),专为足球花式训练设计。融合专业运动商品目录质感和轻微街头风格。哑光高抓地纹理、低弹结构、耐用街头使用质感、黑白配色、几何拼片图案、现代排版、都市细节、可见缝线、真实材质和高级完成度。白色背景居中棚拍、底部柔和阴影、超写实、清晰细节、干净灯光、极简构图,不出现品牌 Logo。

英文提示词

A premium freestyle football ball, slightly smaller than a regular match ball (size 4), designed specifically for football freestyle training. Blend a high-end professional sports catalog aesthetic with a subtle urban street style. Matte high-grip textured surface, low-bounce construction, durable street-ready build, clean black and white color palette, bold geometric panel graphics, sleek modern layout, subtle urban-inspired details, visible stitching, realistic material texture, and a refined premium finish. The design should feel like an original elite sports product, combining the clean sophistication of a top-tier official match ball with the attitude of a street freestyle ball. Centered studio product photography, pure white background, soft shadow underneath, ultra-realistic, sharp details, clean lighting, minimalist composition, high-end commercial product shot, no visible brand logos.

ChatGPT Image 2 生成的花式足球素材

三、巴西街头球场:让场景先稳定下来

第三步生成环境:巴西街头足球场。作者强调真实城市氛围、电影感和落地感,这能让视频不只是“人物在空白背景里动”,而是具备更强的短片质感。

中文整理:白天的巴西城市街头足球场,空场待用。混凝土五人制球场、磨损的划线、金属球门、铁丝围栏、涂鸦墙、巴西街头文化细节、背景中的低矮住宅和建筑、电线杆、热带植物和棕榈树。温暖自然光、柔和阴影、略带日晒褪色的纹理、干净但有生活感,色彩鲜活但写实。整体像真实的巴西街头足球场,电影感、接地气,并带有运动广告式美学。广角构图、平视机位、高细节、超写实、锐利纹理、自然光、无人物。

英文提示词

A realistic urban street football court in Brazil during daytime, empty and ready for play. Concrete futsal-style pitch with worn painted lines, metal goalposts, chain-link fences, graffiti walls, subtle Brazilian street culture details, a few houses and low buildings in the background, utility poles, tropical plants and palm trees, authentic neighborhood atmosphere. Warm natural daylight, soft shadows, slightly sun-faded textures, clean but lived-in look, vibrant yet realistic colors. The scene should feel like a real Brazilian street football playground, cinematic and grounded, with a stylish sports-commercial aesthetic. Wide composition, eye-level camera, highly detailed, ultra-realistic, sharp textures, natural lighting, no people.

ChatGPT Image 2 生成的巴西街头足球场环境

四、动作编舞图:用 16 步动作控制视频节奏

最关键的一步是 choreography sheet。作者把后续动画要执行的动作提前规划成 16 步足球花式动作,让视频模型按动作顺序生成,而不是自由发挥。

中文整理:彩铅手绘风格的足球花式动作分镜信息图,16 个步骤以 4×4 网格排布,每格展示不同足球技巧或控球动作。主体是巴西少年,运动体格、棕褐色皮肤、辫发,穿现代足球花式服装,有巴西街头文化色彩。每帧都与足球互动,动作包括脚尖颠球、around-the-world、交叉步、脚跟挑球、颈部停球、肩部停球、地面控球、旋转控球和花式衔接,并用箭头标注球路、身体移动和脚的位置。顶部标题为“FOOTBALL FREESTYLE — 16 COUNTS — 10 SECONDS — STREET FLOW”。

英文提示词

A colored pencil sketch style choreography sheet infographic for a football freestyle routine. Layout: 16 steps arranged in a clean 4x4 grid, each panel showing a different football trick or control sequence. Subject: a Brazilian teenage boy with an athletic build, tan skin, and braided hair. He has a confident but relaxed street style attitude. He wears a modern football freestyle outfit: oversized jersey, loose shorts, crew socks, and stylish football sneakers. The outfit includes subtle vibrant accents inspired by Brazilian street culture. He looks calm, skillful, and fluid, like a street football prodigy with elite ball control. He is interacting with a football in every frame while performing freestyle movements. Style: hand-drawn colored pencil illustration, soft shading, visible pencil texture, slightly sketchy but clean lines, warm tones mixed with vibrant accents. Movements: each frame shows smooth football freestyle actions such as foot juggling, around-the-world trick, crossover step, heel flick, neck stall, shoulder stall, ground control, spinning control, and stylish freestyle transitions, with small arrows indicating ball direction, body movement flow, and foot placement. Design: modern street football aesthetic, minimal and stylish infographic layout, soft highlights, step numbers (1–16), short captions under each frame describing each move. Text: Title at the top — “FOOTBALL FREESTYLE — 16 COUNTS — 10 SECONDS — STREET FLOW” Environment: outdoor urban football court, street cage or playground, concrete floor, subtle background elements, warm daylight, clean but authentic street atmosphere. Quality: high detail, sharp composition, balanced layout, editorial sports tutorial poster.

足球花式 16 步动作编舞分镜图

五、Seedance 2.0:把所有参考图合成 10 秒视频

最后,作者在 Magnific 中使用 Seedance 2.0,把四张参考图和最终提示词结合起来,生成 10 秒 1080p 视频。这里最值得学习的是:提示词不只是描述画面,还明确规定了动作时间线、全身构图、镜头方式、物理真实感和音乐氛围。

Seedance 2.0 生成足球花式视频关键帧

中文整理:使用 Image1 作为角色参考,Image2 作为足球参考,Image3 作为球场参考,Image4 作为编舞参考。创建一段 10 秒电影感写实足球花式视频,包含 16 个节拍,并按 Image4 中的 16 个动作顺序执行。角色是年轻巴西男性,瘦而运动型,棕褐色皮肤、辫发,穿同样的黄色巴西球衣、蓝色短裤、白袜和白色花式足球鞋。足球使用同样的黑白花式足球。场景使用同样的巴西街头球场。

构图要求:角色必须始终从头到脚完整可见,不要裁切身体;尽量居中,大多数时间面向镜头,保持在球场中央。镜头平滑跟随,并始终让完整人物和足球可见。整体应像一段连续的 freestyle combo,快速、流畅、有连接感,具备电影真实感、自然球体物理、真实身体力学和平滑转场。

10 秒动作时间线

  • 0.0–1.2s:脚尖颠球;around the world
  • 1.2–2.4s:交叉步;脚跟挑球
  • 2.4–3.6s:颈部停球;肩部停球
  • 3.6–4.8s:脚底控球;内侧触球
  • 4.8–6.0s:外侧触球;地面控球
  • 6.0–7.2s:滚拉动作;旋转转身
  • 7.2–8.4s:脚尖停球;跳过足球
  • 8.4–10.0s:绕腿动作;最后用脚踩住足球完成定格
英文提示词

Use @Image1 as the character reference, @Image2 as the ball reference, @Image3 as the court reference, and @Image 4 as the choreography reference.

Create a 10-second cinematic realistic football freestyle video with 16 counts, following the 16 moves from @Image4 in order.

Character: young adult Brazilian male, slim athletic build, tan skin, braided hair, wearing the same yellow Brazil jersey, blue shorts, white socks, and white freestyle sneakers from @Image1 . Keep his identity, outfit, and proportions consistent.

Ball: use the same black-and-white freestyle football from @Image2 .

Location: use the same Brazilian street football court from @Image3 : outdoor concrete futsal court, fences, graffiti, tropical details, warm daylight.

The freestyler must remain fully visible in full body at all times, from head to toe, with no cropped framing. Keep him centered in frame as much as possible, mostly facing the camera, staying near the middle of the court. The camera should follow him smoothly while always keeping his entire body and the ball visible.

Timeline / 16 counts / 10s: 0.0–1.2s toe juggle and around the world; 1.2–2.4s crossover step and heel flick; 2.4–3.6s neck stall and shoulder stall; 3.6–4.8s sole control and inside touch; 4.8–6.0s outside touch and ground control; 6.0–7.2s roll and pull plus spin turn; 7.2–8.4s toe stall and hop over; 8.4–10.0s around the leg and style finish with the ball stopped under his foot.

The routine should feel like one continuous freestyle combo, fast, fluid, and connected. Use cinematic realism, natural ball physics, realistic body mechanics, and smooth transitions.

Camera: mostly wide or medium-wide full-body framing, gentle handheld tracking, minimal reframing, no close-ups that crop the body, no extreme angles, always keep the full character visible.

Music: energetic Brazilian samba rhythm, upbeat percussion, lively street vibe.

这个案例的核心启发

  • 先静态,后动态:先把人物、道具、场景和动作都做成参考素材,再进入视频生成。
  • 把“动作”视觉化:编舞图比纯文字动作描述更容易约束视频模型。
  • 全身构图要写清楚:运动视频最容易裁切身体,所以提示词里反复强调 full body 很重要。
  • 时间线越明确,结果越可控:把 10 秒拆成 8 个时间段和 16 个动作,能显著减少随机性。
  • 适合迁移到其他题材:舞蹈、健身、武术、篮球、产品广告和短剧动作片段都可以套用类似流程。