YouTube 1.8 亿播放的 Cat AI Shorts 复刻全过程

YouTube 1.8 亿播放的 Cat AI Shorts 复刻全过程

从角色三视图到逐场景生成、拼接配乐,我完整复刻了一条 8.9 秒的爆款 Cat AI Shorts(橘猫做菜,1.8 亿次观看)。这篇文章是全部制作过程的真实记录:AI Cat 角色的参考图、提示词、制作方法,都在这里。

创作 AI 视频
VidLux阅读约需 6 分钟

这是 YouTube 上一条 1.8 亿次观看的爆火 Cat AI Shorts:橘猫做饭。它是典型的 AI 生成 Shorts——拟人化动物 + 生活场景 + 快节奏。于是我决定用 VidLux 把它完整复刻一遍。左边是原视频,右边是我们的复刻成品:

原视频:1.8 亿次观看 · 8.9 秒
复刻成品:橘猫做饭 · 4 场景快切 · 9 秒

复刻不是照搬画面,而是把它的制作逻辑拆出来:角色一致性怎么保证、节奏怎么控制、动作怎么做。这篇文章把整个过程拆给你看。

第一步:先造角色——电影级拟人橘猫

复刻的核心是同一只橘猫贯穿 4 个场景(敲鸡、洗鸡、切鸡、烤鸡)。这是 AI 视频最难的环节:角色一致性。

我的做法是:英雄图 → 三视图 → 配场景 → 出视频

先用 GPT Image 2 生成角色英雄图。目标风格是电影级拟人萌宠动画(Photorealistic Anthropomorphic Animal CGI),类似皮克斯/照明工作室那种质感:真实的毛发渲染、次表面散射、电影级打光。而且必须是胖猫(圆肚子是萌点)。这种 Cat AI 角色图,用 VidLux 的图片生成就能直接做出来:

橘猫英雄图:电影级 CGI 拟人胖橘猫
角色英雄图:电影级 CGI 拟人胖橘猫(GPT Image 2)
查看英雄图提示词

Photorealistic anthropomorphic CGI of a very chubby fat orange tabby cat, standing upright on two legs like a human in a natural relaxed pose, facing the camera, paws relaxed at sides. Movie-quality CGI fur rendering with subsurface scattering, every hair strand visible, realistic orange tabby stripes, big round belly, large expressive amber eyes with soft catchlights. Cinematic studio lighting, gentle rim light, shallow depth of field, premium animated movie character look like a Pixar or Illumination film still. Plain neutral gray studio background, full body visible, character design reference, 16:9, no text, no logos, no watermark.

然后以英雄图为参考,生成三视图(正面/侧面/背面在一张图里)。这样三个角度的橘猫来自同一张图,天然是同一个人——比分别生成三个角度再指望"看起来像"可靠得多:

橘猫三视图:正/侧/背
橘猫三视图:同一只猫的正/侧/背三个角度
查看三视图提示词

Using the exact same chubby orange tabby cat from the reference image: photorealistic anthropomorphic CGI character turnaround sheet showing THREE views of the same fat orange tabby cat standing upright - FRONT view, LEFT SIDE view and BACK view, arranged side by side in a row, all three views identical in proportions, fur pattern, chubby body, height and lighting. Movie-quality CGI fur rendering, cinematic lighting, premium animated movie character look, full body, neutral standing pose, paws at sides, centered, evenly spaced, character design sheet, no text, no logos, no watermark.

三视图定稿后,关键就在这:后面 4 个场景的关键帧,都用三视图作为参考输入来生成,保证每个场景里的橘猫都是同一只。

第二步:4 个场景的关键帧

角色定稿后,开始生成 4 个场景的关键帧。每个场景用图生图(image-to-image),参考三视图,提示词描述场景内容。

画幅比例要对齐目标格式:这条 Shorts 是 9:16 竖屏(1080x1920),所以 4 个场景全部按 9:16 生成——从关键帧阶段就对齐,而不是生成后再改比例:

场景 1:橘猫举棒球棒敲鸡
场景 1:举棒球棒敲鸡
场景 2:橘猫洗整鸡
场景 2:水槽洗整鸡
场景 3:橘猫切整鸡
场景 3:切整鸡
场景 4:橘猫给烤鸡刷酱
场景 4:给烤鸡刷酱
查看 4 个场景的完整提示词

场景 1(敲鸡):

Using the exact same chubby orange tabby cat from the reference images: the fat orange tabby cat stands on a dirt path in a sunny countryside field, holding a wooden baseball bat with both paws, ready to swing, a colorful rooster standing beside it looking on, green fields, distant hills, blue sky with white clouds, wooden fence. Photorealistic anthropomorphic animal CGI, movie-quality fur rendering with subsurface scattering, cinematic lighting, premium animated movie character look, vibrant and warm, 9:16 vertical composition, no text, no logos, no watermark.

场景 2(洗鸡):

Using the exact same chubby orange tabby cat from the reference images: the fat orange tabby cat stands at a metal kitchen sink, holding a whole skinned pink-white raw chicken with both paws under running water from the faucet, washing the chicken, water splashing, lemons and tomatoes with red sauce on the counter beside the sink, a red container nearby, bright kitchen background with window, plants and shelves, cozy warm atmosphere. Photorealistic anthropomorphic animal CGI, movie-quality fur rendering, cinematic lighting, vibrant and warm, 9:16 vertical composition, no text, no logos, no watermark.

场景 3(切鸡):

Using the exact same chubby orange tabby cat from the reference images: the fat orange tabby cat wearing a white chef hat and apron stands at the kitchen counter, a WHOLE skinned raw chicken lying on the wooden cutting board in front of it, the cat holding a kitchen knife in one paw raised over the whole chicken ready to cut, lemon and chili peppers around the board, blurred kitchen background with pots and shelves. Photorealistic anthropomorphic animal CGI, movie-quality fur rendering, cinematic lighting, vibrant and warm, 9:16 vertical composition, no text, no logos, no watermark.

场景 4(烤鸡):

Using the exact same chubby orange tabby cat from the reference images: the fat orange tabby cat standing at a grill in the kitchen, holding a blue brush basting sauce over a roasting chicken on the grill with glowing charcoal below, kitchen cabinets and stove in the background. Photorealistic anthropomorphic animal CGI, movie-quality fur rendering, cinematic lighting, vibrant and warm, 9:16 vertical composition, no text, no logos, no watermark.

(4 个场景提示词结构相同:同一只猫 + 场景描述 + 同一段风格后缀。)

第三步:逐场景生成视频

关键帧确认后,用 图生视频 让每个场景动起来。我用的是 PixVerse V6(队列稳定、成本低),每个场景生成 4 秒。图生视频直接上传对应的场景图就够了——三视图的作用到这里结束,它只负责让场景图里的猫保持同一只;从场景图到视频,画面已经定稿,提示词只需要描述动作。

下面是 4 个片段各自的生成过程——输入的关键帧、完整的视频提示词、生成结果:

片段 1:敲鸡

关键帧是乡村田野里举棒准备敲的橘猫,视频提示词要精确到"敲击部位"——用棒头敲公鸡的头,而不是挥棒打棒球:

片段 1 输入关键帧:橘猫举棒球棒
输入关键帧:举棒球棒对准公鸡
片段 1 · PixVerse V6 · 4s · 棒头敲晕公鸡
查看片段 1:敲鸡的视频提示词

The fat orange tabby cat raises the wooden baseball bat high, then strikes the rooster on the head with the thick barrel end of the bat (the bat head), knocking the rooster out. The rooster wobbles and collapses to the ground, comedic effect, dust puff on impact. The bat head hits the rooster, not the bat shaft. Photorealistic anthropomorphic animal CGI, movie-quality fur rendering, cinematic lighting, vibrant and warm, 9:16 vertical, no text, no logos, no watermark.

片段 2:洗鸡

关键帧是水槽里洗剥皮整鸡,视频提示词在"洗"之外加了一个点睛动作——洗完捧起鸡肉亲一下

片段 2 输入关键帧:橘猫洗整鸡
输入关键帧:水槽里洗剥皮整鸡
片段 2 · PixVerse V6 · 4s · 洗整鸡 + 亲吻
查看片段 2:洗鸡的视频提示词

The fat orange tabby cat washes the whole skinned raw chicken under running water in the metal sink, then after washing, picks up the wet chicken with both paws and gives it a loving kiss on the chicken, eyes closed, sweet expression. Photorealistic anthropomorphic animal CGI, movie-quality fur rendering, cinematic lighting, vibrant and warm, 9:16 vertical, no text, no logos, no watermark.

片段 3:切鸡

关键帧是厨师帽橘猫举刀切整鸡,视频提示词在"切"之前加了一个细节——先**"喵"叫一声**再下刀:

片段 3 输入关键帧:橘猫切整鸡
输入关键帧:厨师帽 + 砧板上整鸡
片段 3 · PixVerse V6 · 4s · 喵一声 + 切整鸡
查看片段 3:切鸡的视频提示词

The fat orange tabby cat wearing a white chef hat and apron opens its mouth and lets out a single meow, then raises the kitchen knife and starts chopping the whole raw chicken on the wooden cutting board, cutting into the chicken. Photorealistic anthropomorphic animal CGI, movie-quality fur rendering, cinematic lighting, vibrant and warm, 9:16 vertical, no text, no logos, no watermark.

片段 4:烤鸡

关键帧是橘猫给烤鸡刷酱,视频提示词强调炭火的动态——酱汁刷上去、炭火在下方燃烧:

片段 4 输入关键帧:橘猫给烤鸡刷酱
输入关键帧:烤鸡 + 刷酱
片段 4 · PixVerse V6 · 4s · 烤鸡刷酱
查看片段 4:烤鸡的视频提示词

The fat orange tabby cat brushes sauce over the roasting chicken on the grill, glowing charcoal below, flames licking the chicken. Photorealistic anthropomorphic animal CGI, movie-quality fur rendering, cinematic lighting, vibrant and warm, 9:16 vertical, no text, no logos, no watermark.

一个关键认识:AI 模型对动作的理解依赖提示词的精确程度。 动作要写清"做什么、用什么部位、结果怎样",比如敲鸡要写"用棒头敲公鸡的头、公鸡晕倒",洗鸡要写"剥皮的整鸡、洗完亲一下"——少一个细节,模型就按自己的理解发挥。

第四步:拼接成片 + 配乐

4 段视频各 4 秒,但原视频是 8.9 秒、每场景 2.2 秒。原视频的节奏是加速过的——每个动作压缩到 2.2 秒内完成,非常紧凑。

所以我在剪辑时做了两件事:

  1. 加速:把每段视频加速到原来的 1.8 倍(4 秒变 2.2 秒),动作保持完整,但节奏更紧凑。剪映、CapCut 这类剪辑软件里都有"变速"功能,拖一下就行。
  2. 拼接:把加速后的 4 段按顺序接在一起,变成一个连续的视频。

最后配乐:原视频是纯背景音乐,我配了同款轻快曲目。拼接 + 配乐后就是开头那个成品。

复刻的完整参数对照:

项目原视频复刻版
时长8.9s9.0s
画幅1080x1920 (9:16)720x1280 (9:16)
节奏4 场景 × 2.2s 快切4 场景 × 2.2s 快切(1.8x 加速)
角色橘猫拟人电影级 CGI 橘猫(同一只)
音频轻快 BGM同款曲目

复盘:复刻爆款的核心是什么

做完这条视频,我最大的体会是:复刻的难点不在生成,在分析和设计。

  1. 先拆解再动手:把原视频的节奏、内容、风格全部量化,生成才有依据
  2. 角色一致性是地基:英雄图 → 三视图 → 场景图 → 视频,每一环都用上一环做参考,4 个场景才是同一只猫——观众不会出戏
  3. 动作要精确到部位:敲鸡和挥棒、棒头和棒身、洗鸡和洗菜——提示词里差一个词,画面就完全不同
  4. 对齐画幅和节奏:9:16 竖屏从关键帧阶段就要对齐;快节奏靠加速实现,不是让模型"快点动"

说到底,AI 是画师,你是导演。 原视频的爆款逻辑(拟人化动物 + 反差萌 + 快节奏)是分析出来的,角色的设定、场景的顺序、动作的细节,都是人做的判断。AI 负责把判断变成画面。AI 猫视频(AI Cat video)并不难做——难的是像这条一样,把一只猫的角色一致性做到 4 个场景不穿帮。

想自己试一条?VidLux 的图生视频就能做。先从一张角色图开始,按这套流程走一遍,你也能复刻出属于自己的 Cat AI 爆款 Shorts