ai-video-platform

name	ai-video-platform
description	构建 AI 短视频一键生产发布客户端。整合 MoneyPrinterTurbo（AI视频生成）与 AiToEarn/social-auto-upload（多平台一键发布）。支持5种视频生成方式（拼素材/可灵/即梦/Wan2.1/Runway），AI文案生成，自动配音字幕，一键发布到抖音/快手/B站/小红书/微博/视频号/TikTok/YouTube/Instagram等12+平台。技术栈：Electron + React + Python FastAPI + Playwright + SQLite。触发词：AI视频平台、视频生成发布、短视频客户端、MoneyPrinterTurbo、AiToEarn、一键发布、视频矩阵。

AI 短视频一键生产发布平台 SKILL

项目目标

将 MoneyPrinterTurbo（视频生成）和 AiToEarn / social-auto-upload（多平台发布）整合为一个 Windows 桌面客户端。用户输入主题 → AI生成视频 → 一键发布全平台。

---

核心依赖项目

项目	用途	地址
MoneyPrinterTurbo	视频生成核心（文案/素材/配音/字幕/合成）	https://github.com/harry0703/MoneyPrinterTurbo
social-auto-upload	多平台发布核心（抖音/B站/小红书/快手/视频号/TikTok）	https://github.com/dreammis/social-auto-upload
AiToEarn	Electron架构参考 + 发布平台参考	https://github.com/yikart/AiToEarn

---

完整目录结构

ai-video-platform/
├── electron/
│   ├── main.ts                  # 主进程，窗口管理，启动Python后端
│   ├── preload.ts               # 预加载，暴露IPC给渲染进程
│   └── python-manager.ts       # Python子进程管理
├── src/                         # React前端
│   ├── App.tsx
│   ├── main.tsx
│   ├── pages/
│   │   ├── Create.tsx           # 视频创作页（核心）
│   │   ├── Publish.tsx          # 发布管理页
│   │   ├── Queue.tsx            # 任务队列页
│   │   ├── Accounts.tsx         # 账号管理页
│   │   ├── Library.tsx          # 素材库页
│   │   └── Settings.tsx         # 设置页
│   ├── components/
│   │   ├── VideoPreview.tsx     # 视频预览组件
│   │   ├── CopywriterPanel.tsx  # 文案生成面板
│   │   ├── PublishPanel.tsx     # 发布配置面板
│   │   └── TaskCard.tsx         # 任务卡片组件
│   ├── store/
│   │   ├── videoStore.ts        # 视频生成状态
│   │   ├── publishStore.ts      # 发布状态
│   │   └── settingsStore.ts     # 设置状态
│   └── api/
│       └── client.ts            # 封装后端API调用
├── backend/                     # Python FastAPI后端
│   ├── main.py                  # FastAPI入口
│   ├── config.py                # 配置管理（API Keys等）
│   ├── video/
│   │   ├── generator.py         # 视频生成调度器（路由到不同方式）
│   │   ├── stock.py             # 拼素材方式（Pexels/Pixabay）
│   │   ├── kling.py             # 可灵API
│   │   ├── jimeng.py            # 即梦API
│   │   ├── wan21.py             # Wan2.1本地模型
│   │   └── runway.py            # Runway API
│   ├── llm/
│   │   └── copywriter.py        # AI文案生成（支持多模型）
│   ├── tts/
│   │   └── synthesizer.py       # 配音合成（Edge/Azure/OpenAI TTS）
│   ├── subtitle/
│   │   └── generator.py         # 字幕生成（faster-whisper）
│   ├── compose/
│   │   └── merger.py            # 视频合成（moviepy+ffmpeg）
│   ├── publish/
│   │   ├── base.py              # 发布基类
│   │   ├── douyin.py            # 抖音发布
│   │   ├── bilibili.py          # B站发布
│   │   ├── xiaohongshu.py       # 小红书发布
│   │   ├── kuaishou.py          # 快手发布
│   │   ├── wechat_video.py      # 微信视频号发布
│   │   ├── weibo.py             # 微博发布
│   │   ├── tiktok.py            # TikTok发布
│   │   └── youtube.py           # YouTube发布
│   ├── db/
│   │   ├── database.py          # SQLite连接
│   │   └── models.py            # 数据模型
│   └── vendor/                  # 第三方项目（git clone进来）
│       ├── MoneyPrinterTurbo/
│       └── social-auto-upload/
├── accounts/                    # 平台账号Cookie存储（加密）
├── output/                      # 生成视频输出目录
├── assets/                      # 本地素材库
├── package.json
├── vite.config.ts
├── tsconfig.json
├── electron-builder.yml
└── requirements.txt

---

Step 1: 初始化项目

mkdir ai-video-platform && cd ai-video-platform

# 初始化 npm
npm init -y

# 安装 Electron + React + UI
npm install electron react react-dom antd @ant-design/icons zustand react-router-dom
npm install -D typescript @types/react @types/react-dom vite @vitejs/plugin-react electron-builder

# 安装 Python 依赖
pip install fastapi uvicorn moviepy edge-tts faster-whisper requests pyyaml openai playwright
pip install --break-system-packages yt-dlp  # 如果系统包冲突

# 安装 Playwright 浏览器
python -m playwright install chromium

# 克隆核心依赖
git clone https://github.com/harry0703/MoneyPrinterTurbo backend/vendor/MoneyPrinterTurbo
git clone https://github.com/dreammis/social-auto-upload backend/vendor/social-auto-upload

---

Step 2: Python 后端

backend/main.py

from fastapi import FastAPI
from fastapi.middleware.cors import CORSMiddleware
from video.generator import router as video_router
from llm.copywriter import router as llm_router
from tts.synthesizer import router as tts_router
from publish.douyin import router as douyin_router
from publish.bilibili import router as bili_router
from publish.xiaohongshu import router as xhs_router
from publish.kuaishou import router as ks_router
from db.database import init_db

app = FastAPI(title="AI Video Platform API")
app.add_middleware(CORSMiddleware, allow_origins=["*"], allow_methods=["*"], allow_headers=["*"])

@app.on_event("startup")
async def startup():
    init_db()

app.include_router(video_router, prefix="/api/video", tags=["video"])
app.include_router(llm_router, prefix="/api/llm", tags=["llm"])
app.include_router(tts_router, prefix="/api/tts", tags=["tts"])
app.include_router(douyin_router, prefix="/api/publish/douyin", tags=["publish"])
app.include_router(bili_router, prefix="/api/publish/bilibili", tags=["publish"])
app.include_router(xhs_router, prefix="/api/publish/xiaohongshu", tags=["publish"])
app.include_router(ks_router, prefix="/api/publish/kuaishou", tags=["publish"])

backend/video/generator.py（调度器）

from fastapi import APIRouter
from pydantic import BaseModel
from typing import Literal
import asyncio, uuid

router = APIRouter()
tasks = {}  # 内存任务状态（生产环境用 DB）

class VideoRequest(BaseModel):
    topic: str
    mode: Literal["stock", "kling", "jimeng", "wan21", "runway"]
    aspect_ratio: Literal["9:16", "16:9"] = "9:16"
    duration: int = 30  # 秒
    script: str = ""    # 可选，传入则跳过AI生成文案

@router.post("/generate")
async def generate_video(req: VideoRequest):
    task_id = str(uuid.uuid4())
    tasks[task_id] = {"status": "pending", "progress": 0, "url": None}
    asyncio.create_task(_run_generation(task_id, req))
    return {"task_id": task_id}

@router.get("/status/{task_id}")
async def get_status(task_id: str):
    return tasks.get(task_id, {"status": "not_found"})

async def _run_generation(task_id: str, req: VideoRequest):
    tasks[task_id]["status"] = "running"
    try:
        if req.mode == "stock":
            from .stock import generate
        elif req.mode == "kling":
            from .kling import generate
        elif req.mode == "jimeng":
            from .jimeng import generate
        elif req.mode == "wan21":
            from .wan21 import generate
        elif req.mode == "runway":
            from .runway import generate
        output_path = await generate(req)
        tasks[task_id] = {"status": "done", "progress": 100, "url": output_path}
    except Exception as e:
        tasks[task_id] = {"status": "error", "error": str(e)}

backend/video/stock.py（拼素材，基于MoneyPrinterTurbo）

import sys, os
sys.path.insert(0, os.path.join(os.path.dirname(__file__), '../vendor/MoneyPrinterTurbo'))
from app.services.video import VideoService
from app.services.material import MaterialService

async def generate(req) -> str:
    """复用 MoneyPrinterTurbo 的核心逻辑生成拼素材视频"""
    # 1. 用 MoneyPrinterTurbo 的 MaterialService 搜索素材
    materials = MaterialService.search_videos(req.topic, count=5)
    # 2. 用 VideoService 合成视频
    output = VideoService.combine(
        materials=materials,
        script=req.script,
        aspect_ratio=req.aspect_ratio,
        output_dir="output/"
    )
    return output

backend/video/kling.py（可灵API）

import requests, time

KLING_BASE = "https://api.klingai.com/v1"

def

### backend/video/kling.py
import requests, time, jwt

KLING_BASE = 'https://api.klingai.com/v1'

def _get_token(ak, sk):
    payload = {'iss': ak, 'exp': int(time.time())+1800, 'nbf': int(time.time())-5}
    return jwt.encode(payload, sk, algorithm='HS256')

async def generate(req):
    from config import get_config
    cfg = get_config()
    token = _get_token(cfg.kling_ak, cfg.kling_sk)
    headers = {'Authorization': f'Bearer {token}'}
    body = {'model_name': 'kling-v1', 'prompt': req.script or req.topic,
            'aspect_ratio': req.aspect_ratio, 'duration': str(min(req.duration,10))}
    r = requests.post(f'{KLING_BASE}/videos/text2video', headers=headers, json=body)
    task_id = r.json()['data']['task_id']
    for _ in range(120):
        time.sleep(5)
        r = requests.get(f'{KLING_BASE}/videos/text2video/{task_id}', headers=headers)
        d = r.json()['data']
        if d['task_status'] == 'succeed': return d['task_result']['videos'][0]['url']
        if d['task_status'] == 'failed': raise Exception('Kling failed')
    raise TimeoutError('Kling timeout')

### backend/video/jimeng.py
async def generate(req):
    from config import get_config
    import requests, time
    cfg = get_config()
    headers = {'Authorization': f'Bearer {cfg.jimeng_api_key}'}
    body = {'prompt': req.script or req.topic, 'ratio': req.aspect_ratio.replace(':','_')}
    r = requests.post('https://jimeng.jianying.com/mweb/v1/generate_video', headers=headers, json=body)
    task_id = r.json()['data']['task_id']
    for _ in range(120):
        time.sleep(5)
        r = requests.get(f'https://jimeng.jianying.com/mweb/v1/task/{task_id}', headers=headers)
        d = r.json()['data']
        if d['status'] == 'done': return d['video_url']
    raise TimeoutError('Jimeng timeout')

### backend/video/runway.py
async def generate(req):
    import runwayml, time
    from config import get_config
    cfg = get_config()
    client = runwayml.RunwayML(api_key=cfg.runway_api_key)
    ratio = '768:1280' if req.aspect_ratio == '9:16' else '1280:768'
    task = client.image_to_video.create(model='gen4_turbo', prompt_text=req.script or req.topic, ratio=ratio)
    for _ in range(120):
        time.sleep(5)
        result = client.tasks.retrieve(task.id)
        if result.status == 'SUCCEEDED': return result.output[0]
        if result.status == 'FAILED': raise Exception('Runway failed')
    raise TimeoutError('Runway timeout')

### backend/video/wan21.py
async def generate(req):
    from diffusers import AutoPipelineForText2Video
    import torch
    from config import get_config
    cfg = get_config()
    pipe = AutoPipelineForText2Video.from_pretrained(cfg.wan21_model_path, torch_dtype=torch.bfloat16).to('cuda')
    output = pipe(prompt=req.script or req.topic, num_frames=req.duration*16).frames[0]
    out_path = f'output/wan21_{req.topic[:10]}.mp4'
    output.save(out_path)
    return out_path

---

## Step 3: AI文案生成

PROMPT_TEMPLATE:
  主题:{topic} 平台:{platform} 风格:{style}
  要求: 1.开头钩子 2.口语化 3.结尾引导 4.生成{count}个版本用---分隔

模型支持: DeepSeek(默认)/OpenAI/Gemini/通义千问
配置: llm_api_key + llm_base_url + llm_model

---

## Step 4: 配音(Edge TTS)

音色选项:
  zh-female: zh-CN-XiaoxiaoNeural
  zh-male: zh-CN-YunxiNeural
  en-female: en-US-JennyNeural
  en-male: en-US-GuyNeural

调用: edge_tts.Communicate(text, voice).save(output_path)

---

## Step 5: 字幕(faster-whisper)

model = WhisperModel('base', device='cpu', compute_type='int8')
segments, _ = model.transcribe(audio_path, language='zh')
返回: [{start, end, text}, ...]

---

## Step 6: 发布模块

基于 social-auto-upload 直接复用各平台 uploader:
  douyin: uploader.douyin_uploader.main.DouYinVideo
  bilibili: uploader.bilibili_uploader.main.BilibiliUploader
  xiaohongshu: uploader.xhs_uploader.main.XhsUploader
  kuaishou: uploader.ks_uploader.main.KSVideo
  tiktok: uploader.tiktok_uploader.main.TiktokVideo

账号Cookie存储路径: accounts/{platform}_{account_name}.json
首次登录: 运行 setup 函数打开浏览器扫码/登录

---

## Step 7: Electron主进程

启动流程:
  1. app.whenReady() -> startPythonBackend()
  2. Python: uvicorn main:app --host 127.0.0.1 --port 8765
  3. 等待后端就绪(轮询 /health 接口)
  4. 加载前端页面

前后端通信: HTTP REST (http://127.0.0.1:8765)

---

## Step 8: 配置项(config.json)

LLM: llm_api_key, llm_base_url, llm_model
视频生成: kling_ak, kling_sk, jimeng_api_key, runway_api_key, wan21_model_path
素材: pexels_api_key, pixabay_api_key
代理: proxy (海外平台用)

---

## Step 9: 打包

Python后端: pyinstaller --onefile backend/main.py -> dist/backend_server.exe
Electron: electron-builder --win -> dist/AI视频平台 Setup.exe
extraResources: backend/dist/, accounts/, output/

---

## 完整API清单

POST /api/llm/copywrite         - 生成文案
POST /api/video/generate         - 提交视频生成任务
GET  /api/video/status/{id}      - 查询进度
POST /api/tts/synthesize         - 合成配音
POST /api/subtitle/generate      - 生成字幕
POST /api/publish/douyin/publish - 发布抖音
POST /api/publish/bilibili/publish - 发布B站
POST /api/publish/xiaohongshu/publish - 发布小红书
POST /api/publish/kuaishou/publish - 发布快手
POST /api/publish/tiktok/publish - 发布TikTok
POST /api/publish/youtube/publish - 发布YouTube
GET  /api/accounts/list          - 获取账号列表
GET  /api/queue/list             - 获取任务队列
POST /api/settings/save          - 保存配置

---

## 注意事项

1. Python后端端口 8765，Electron启动时自动拉起，退出时自动关闭
2. 账号Cookie加密存储，不上传
3. 可灵/即梦/Runway均为异步任务，轮询间隔5秒，超时10分钟
4. social-auto-upload需要Playwright+Chrome，首次运行自动安装
5. Wan2.1需要CUDA GPU，无GPU时前端自动禁用该选项
6. 海外平台(TikTok/YouTube/Instagram)需配置代理
7. 拼素材需要Pexels API Key(免费注册)