# I Built an AI Anime Factory That Runs While I Sleep
**作者**: fokki
**日期**: 2026-05-18T09:43:33.000Z
**来源**: [https://x.com/0x_fokki/status/2056309709720707086](https://x.com/0x_fokki/status/2056309709720707086)
---
$8,217 last month. I touched it maybe 3 hours total. Here's every piece of the system.
I got obsessed with one question six months ago.
What if anime content - characters, voice, music, story - could be generated faster than anyone could watch it?
Turns out it can. And YouTube pays you for every hour they watch.
The factory runs 24/7. I sleep. It publishes.
Here's the full build.
## What the Factory Actually Produces
Three types of content, all automated:
> → Anime story shorts - 8–15 minute episodes with AI characters, AI voiceover, AI music
> → Lofi anime channels -8-hour study streams, looping visuals, background music
> → Anime OST channels - AI-generated soundtracks uploaded as "official" channel music
Each type monetizes differently. Together they hit $8,000+/month.
One pipeline builds all three.
## The Pipeline
Claude → Midjourney → Runway → ElevenLabs → Suno → Make
script → characters → animate → voice → music → publish
Six tools. Fully automated after setup.
## Step 1 - Claude Writes the Script
Every piece of content starts here. Claude writes the episode script, the scene descriptions for Midjourney, and the voiceover lines for ElevenLabs.
Prompt for episode script:
```
You are a writer for a dark fantasy anime series.
Write episode 3 of a 12-episode series called "Shadow Protocol."
The main character is Kaito, 17, a hacker who discovers
he can manipulate digital reality with his mind.
Episode 3: Kaito infiltrates a corporate server and finds
evidence that the company has been running experiments
on people with abilities like his.
Format the output as:
- Scene description (for image generation)
- Character dialogue
- Narrator lines
- Emotional tone notes
Each scene should be 30-45 seconds when read aloud.
Total episode: 12 minutes.
Dark tone. Fast pacing. Cliffhanger ending.
```
Prompt for scene descriptions:
```
Convert this script scene into a Midjourney image prompt.
Scene: Kaito sits alone in a dark server room,
blue holographic screens surrounding him,
lines of code reflected in his glasses,
tension on his face as he reads classified files.
Output format:
- Main subject description
- Environment and lighting
- Art style (anime, cinematic)
- Mood and color palette
Make it visually dramatic and cinematic.
```
## Step 2 - Midjourney Generates the Characters
Base character prompt:
```
anime teenage boy, dark messy hair, sharp eyes,
black hoodie with hood down, fingerless gloves,
sitting in dark server room, surrounded by blue holographic screens,
code and data streams reflected in his glasses,
tense focused expression, reading something alarming,
dramatic side lighting, deep shadows,
cinematic anime style, ultra detailed,
Makoto Shinkai inspired, 4k
```
Action scenes:
```
same character mid-motion, dynamic camera angle,
speed lines, dramatic lighting shift from blue to red,
high contrast, motion blur on background,
anime fight sequence composition,
Studio Trigger style, explosive energy
```
Emotional scenes:
```
same character close-up, rain outside window,
soft diffused light, reflective mood,
slight downward gaze, complex emotion,
watercolor-influenced backgrounds,
Your Name aesthetic
```
Generate 4 variations per scene. Upscale the best. You need 15–20 images per episode.
## Step 3 - Runway Animates Everything
Action scene:
```
Camera slowly pushes in toward the character.
Holographic screens flicker and pulse with data.
Character's eyes move across the screen reading fast.
Subtle breathing movement in the chest.
Blue light dances across his face.
5 seconds. Tense, urgent energy.
```
Emotional scene:
```
Rain streaks down the window behind the character.
Soft light shifts slightly as clouds pass.
Character exhales slowly, visible in cold air.
Hair moves gently from air conditioning.
Camera barely drifts left over 6 seconds.
Melancholic. Still. Cinematic.
```
Establishing shot:
```
Wide shot of a futuristic city at night.
Neon lights reflect in wet streets below.
Flying vehicles pass in the distance.
Camera pulls back slowly revealing scale.
Fog rolls between buildings.
Blade Runner meets Ghost in the Shell aesthetic.
8 seconds.
```
## Step 4 - ElevenLabs Does the Voices
Best voices for anime content:
- Kaito (main): Adam - young, intense, slight edge
- Narrator: Antoni - deep, cinematic, authoritative
- Female lead: Bella - warm but with mystery
Settings:
> Stability: 0.35
> Similarity: 0.85
> Style exaggeration: 0.40
> Speaker boost: ON
Directing the performance:
```
Read this line with urgency and quiet fear,
like you just discovered something you weren't supposed to see
and someone might be listening:
"These files... they've been watching us for years."
Pause 0.8 seconds after "files."
Barely above a whisper on the last four words.
```
ElevenLabs follows emotional direction when you write it into the prompt. Most people just paste text. Write the performance.
## Step 5 - Suno Generates the Soundtrack
Opening theme:
```
Anime opening theme, J-pop influenced,
energetic but with dark undertones,
electric guitar lead melody, synth layers,
taiko drums mixed with electronic beats,
builds from quiet intro to explosive chorus at 0:45,
male vocal melody, no lyrics, just melodic "ah" sounds,
runtime 1:30, feels like Attack on Titan meets Demon Slayer,
cinematic, hype, memorable hook
```
Background tension:
```
Cinematic dark ambient, anime thriller style,
low cello drones, distant piano notes,
subtle electronic glitch sounds every 8-12 seconds,
builds tension slowly over 3 minutes,
never resolves, always feels like something is coming,
Hans Zimmer meets anime OST,
instrumental only, loop-friendly
```
Emotional ending:
```
Soft melancholic piano, anime ED style,
slow tempo 68 BPM,
acoustic guitar joins at 0:30,
violin enters at 1:00 adding emotional weight,
feels like the end of an episode where everything changed,
Your Name energy,
3 minutes, fades out gently
```
All tracks are yours. Commercial license on the $8/month plan.
## Step 6 - Make Automates Everything
The scenario runs every Tuesday and Friday at 9am:
> Trigger: Schedule (2x per week)
> 1. Pull new script from Google Drive
> 2. Send scene list to Midjourney via API
> 3. Download generated images to Drive
> 4. Send dialogue to ElevenLabs
> 5. Download audio files
> 6. Combine in CapCut template
> 7. Upload to YouTube with title and description
> 8. Post to X with clip and article link
> 9. Send Telegram notification with stats
Setup time: 4 hours once. Runs forever after.
## How You Make Money
YouTube AdSense
Anime content RPM: $3–$8 per 1,000 views. At 800k monthly views: $2,400–$6,400/month.
Channel memberships
Early episode access at $4.99/month. 200 members = $1,000/month passive.
Music licensing
Your Suno tracks on Musicbed, Artlist, Gumroad. "AI Anime OST Pack, 30 tracks" at $19. Sell it forever.
Sponsorships
Gaming apps, VPN services, anime merch brands. At 30k subscribers: $300–$800 per integration.
## The Numbers
> Month 1: setup, $0
> Month 2: $400-800
> Month 3: $1,500-2,500
> Month 6: $4,000-6,000
> Month 9: $6,000-9,000
> Month 12: $8,000-12,000+
The factory runs. You direct it.
Your only job: pick the story, pick the aesthetic, approve the output.
Six tools. One pipeline. Runs while you sleep.
Save this. Build it this weekend.
part 2 covers the monetization stack i didn't publish here.
comment "BUILD" and i'll send it directly.
won't post it publicly.
## 相关链接
- [fokki](https://x.com/0x_fokki)
- [@0x_fokki](https://x.com/0x_fokki)
- [13K](https://x.com/0x_fokki/status/2056309709720707086/analytics)
- [Upgrade to Premium](https://x.com/i/premium_sign_up)
- [5:43 PM · May 18, 2026](https://x.com/0x_fokki/status/2056309709720707086)
- [13.6K Views](https://x.com/0x_fokki/status/2056309709720707086/analytics)
- [View quotes](https://x.com/0x_fokki/status/2056309709720707086/quotes)
---
*导出时间: 2026/5/19 12:45:43*
---
## 中文翻译
# 我建了一个在我睡觉时自动运行的动漫 AI 工厂
**作者**: fokki
**日期**: 2026-05-18T09:43:33.000Z
**来源**: [https://x.com/0x_fokki/status/2056309709720707086](https://x.com/0x_fokki/status/2056309709720707086)
---
上个月赚了 8,217 美元。我总共只花了大概 3 小时打理它。以下是这套系统的每一个细节。
六个月前，我对一个问题着了迷：
如果动漫内容——角色、配音、音乐、剧情——生成的速度快于任何人观看的速度，会怎样？
事实证明这是可行的。而且 YouTube 会为你支付的每一小时观看时长付费。
工厂全天候 24/7 运行。我睡觉。它发布。
以下是完整的构建过程。
## 工厂实际上生产什么
三种类型的内容，全部自动化：
> → 动漫故事短片——8-15 分钟的剧集，包含 AI 角色、AI 配音、AI 音乐
> → Lofi 动漫频道——8 小时的学习直播流，循环视觉效果，背景音乐
> → 动漫 OST 频道——AI 生成的原声带，上传为“官方”频道音乐
每种类型的变现方式不同。加起来它们实现了每月 8,000+ 美元的收入。
一条流水线就能构建这三种内容。
## 流水线
Claude → Midjourney → Runway → ElevenLabs → Suno → Make
剧本 → 角色 → 动画 → 配音 → 音乐 → 发布
六个工具。设置完成后全自动运行。
## 步骤 1 - Claude 编写剧本
每一段内容都从这里开始。Claude 编写剧集剧本、给 Midjourney 的场景描述，以及给 ElevenLabs 的配音台词。
剧集剧本提示词：
```
你是一部黑暗奇幻动漫系列的编剧。
编写一部名为“暗影协议”的 12 集系列剧的第 3 集。
主角是 Kaito，17 岁，一名黑客，他发现自己
可以用思想操纵数字现实。
第 3 集：Kaito 潜入公司服务器，发现
证据表明该公司一直在对拥有像他这样能力的人进行实验。
将输出格式化为：
- 场景描述（用于图像生成）
- 角色对话
- 旁白台词
- 情感基调注释
每个场景朗读时应为 30-45 秒。
整集时长：12 分钟。
黑暗基调。快节奏。悬念结尾。
```
场景描述提示词：
```
将此剧本场景转换为 Midjourney 图像提示词。
场景：Kaito 独自坐在黑暗的服务器机房里，
蓝色的全息屏幕环绕着他，
一行行代码反射在他的眼镜里，
阅读机密文件时脸上充满紧张感。
输出格式：
- 主体描述
- 环境和光照
- 艺术风格（动漫、电影感）
- 情绪和调色板
使其在视觉上具有戏剧性和电影感。
```
## 步骤 2 - Midjourney 生成角色
基础角色提示词：
```
动漫少年，黑色凌乱头发，眼神锐利，
黑色连帽衫（帽子未戴上），露指手套，
坐在黑暗的服务器机房里，被蓝色全息屏幕包围，
代码和数据流反射在他的眼镜中，
紧张专注的表情，正在阅读令人震惊的信息，
戏剧性的侧光，深邃的阴影，
电影级动漫风格，超精细，
灵感来自 Makoto Shinkai（新海诚），4k
```
动作场景：
```
同一角色动作中，动态相机角度，
速度线，光线从蓝色戏剧性地转变为红色，
高对比度，背景运动模糊，
动漫战斗序列构图，
Trigger 风格，爆发性能量
```
情感场景：
```
同一角色特写，窗外下雨，
柔和的漫射光，反思的情绪，
轻微的低头注视，复杂的情感，
水彩风格背景，
《你的名字》美学
```
每个场景生成 4 个变体。放大最好的。每集你需要 15-20 张图像。
## 步骤 3 - Runway 制作动画
动作场景：
```
相机缓慢推向角色。
全息屏幕闪烁并脉动数据。
角色的眼睛快速移动阅读屏幕内容。
胸部有微妙的呼吸动作。
蓝光在他的脸上舞动。
5 秒。紧张、急促的能量。
```
情感场景：
```
雨点在角色身后的窗户上划过。
随着云层飘过，柔和的光线略有变化。
角色缓慢呼气，在冷空气中清晰可见。
头发因空调风而轻轻飘动。
相机在 6 秒内几乎难以察觉地向左漂移。
忧郁。静止。电影感。
```
定场镜头：
```
未来城市的夜景广角镜头。
霓虹灯反射在下方潮湿的街道上。
飞行车辆在远处经过。
相机缓慢拉后展示规模。
雾气在建筑物间翻滚。
《银翼杀手》遇上《攻壳机动队》的美学。
8 秒。
```
## 步骤 4 - ElevenLabs 制作配音
最适合动漫内容的声音：
- Kaito (主角): Adam - 年轻、强烈、略带锋芒
- Narrator (旁白): Antoni - 深沉、电影感、权威
- Female lead (女主): Bella - 温暖但带有神秘感
设置：
> Stability (稳定性): 0.35
> Similarity (相似度): 0.85
> Style exaggeration (风格夸张度): 0.40
> Speaker boost (扬声器增强): ON
指导表演：
```
带着紧迫感和无声的恐惧朗读这句台词，
就像你刚刚发现了你不该看到的东西，
而且可能有人在监听：
“这些文件……他们已经监视我们好几年了。”
在“文件”之后停顿 0.8 秒。
最后四个词几乎是耳语。
```
当你把情感指导写进提示词时，ElevenLabs 会遵循它。大多数人只是粘贴文本。要写出表演细节。
## 步骤 5 - Suno 生成原声带
片头曲：
```
动漫片头曲，受 J-Pop 影响，
充满活力但带有黑暗底色，
电吉他主旋律，合成器层，
太鼓混合电子节拍，
从安静的前奏逐渐构建，在 0:45 达到爆发性副歌，
男性声乐旋律，无歌词，只有旋律性的“啊”声，
时长 1:30，感觉像《进击的巨人》遇上《鬼灭之刃》，
电影感，热血，令人难忘的记忆点
```
背景张力：
```
电影黑暗环境音，动漫惊悚风格，
低沉的大提琴长音，遥远的钢琴音符，
每 8-12 秒有微妙的电子故障音，
在 3 分钟内缓慢建立张力，
从不解决，总是感觉有什么东西要来了，
Hans Zimmer（汉斯·季默）遇上动漫 OST，
纯器乐，适合循环
```
情感结局：
```
柔和的忧郁钢琴，动漫 ED 风格，
慢节奏 68 BPM，
原声吉他于 0:30 加入，
小提琴于 1:00 进入，增加情感分量，
感觉像是一切都改变的那一集的结尾，
《你的名字》那种氛围，
3 分钟，轻轻淡出
```
所有曲目都属于你。每月 8 美元的套餐包含商业许可。
## 步骤 6 - Make 自动化一切
该场景每周二和周五上午 9 点运行：
> 触发器：计划（每周 2 次）
> 1. 从 Google Drive 获取新剧本
> 2. 通过 API 将场景列表发送给 Midjourney
> 3. 将生成的图像下载到 Drive
> 4. 将对话发送给 ElevenLabs
> 5. 下载音频文件
> 6. 在 CapCut 模板中合成
> 7. 上传到 YouTube 并附上标题和描述
> 8. 发布到 X（推特），附上片段和文章链接
> 9. 发送带有统计数据的 Telegram 通知
设置时间：一次性 4 小时。之后永久运行。
## 你如何赚钱
YouTube AdSense
动漫内容 RPM：每 1,000 次浏览 3-8 美元。每月 80 万浏览量：2,400-6,400 美元/月。
频道会员
每集抢先看，4.99 美元/月。200 名会员 = 1,000 美元/月被动收入。
音乐授权
你的 Suno 曲目放在 Musicbed、Artlist、Gumroad 上销售。“AI Anime OST Pack, 30 tracks”售价 19 美元。永久销售。
赞助
游戏应用、VPN 服务、动漫周边品牌。3 万订阅者：每次植入整合收费 300-800 美元。
## 收入数据
> 第 1 个月：设置，0 美元
> 第 2 个月：400-800 美元
> 第 3 个月：1,500-2,500 美元
> 第 6 个月：4,000-6,000 美元
> 第 9 个月：6,000-9,000 美元
> 第 12 个月：8,000-12,000+ 美元
工厂运行。你指挥它。
你唯一的工作：选择故事，选择美学，批准输出。
六个工具。一条流水线。在你睡觉时运行。
保存这篇文章。这个周末把它建出来。
第二部分涵盖了我这里未发布的变现堆栈。
评论“BUILD”，我会直接发给你。
不会公开发布。
## 相关链接
- [fokki](https://x.com/0x_fokki)
- [@0x_fokki](https://x.com/0x_fokki)
- [13K](https://x.com/0x_fokki/status/2056309709720707086/analytics)
- [升级到 Premium](https://x.com/i/premium_sign_up)
- [5:43 PM · May 18, 2026](https://x.com/0x_fokki/status/2056309709720707086)
- [13.6K Views](https://x.com/0x_fokki/status/2056309709720707086/analytics)
- [View quotes](https://x.com/0x_fokki/status/2056309709720707086/quotes)
---
*导出时间: 2026/5/19 12:45:43*