Faceless AI

Create viral TikTok & YouTube videos with AI. No camera, no editing, just results.

Features

• 🎯 AI Script Writing - Hook-driven scripts optimized for engagement (GPT-4o-mini)
• 🎨 AI Image Generation - Stunning visuals with Flux Schnell (9:16 vertical)
• 🎙️ AI Voiceover - Natural-sounding voices with OpenAI TTS (6 voices)
• 🎬 Auto Video Assembly - FFmpeg-based stitching with captions
• ⚡ Fast - From idea to video in under 2 minutes

Tech Stack

• Frontend: Next.js 14 (App Router), React, Tailwind CSS
• AI Models: OpenAI GPT-4o-mini + TTS, Replicate Flux Schnell
• Video: FFmpeg for assembly, ASS subtitles for captions
• State: In-memory (swap for Redis/DB in production)

Getting Started

1. Clone & Install

git clone <repo>
cd storyshort-clone
npm install

2. Environment Setup

cp .env.local.example .env.local

Edit .env.local and add your API keys:

OPENAI_API_KEY=sk-...
REPLICATE_API_TOKEN=r8_...

Don't have API keys? The app works in demo mode with mock data!

3. Install FFmpeg (required for video assembly)

# macOS
brew install ffmpeg

# Ubuntu/Debian
sudo apt install ffmpeg

# Windows (via chocolatey)
choco install ffmpeg

4. Run

npm run dev

Open http://localhost:3000 🚀

How It Works

1. Enter topic → "The mysterious disappearance of the Mayan civilization"
2. Pick style → Documentary, Horror, Anime, Fantasy, etc.
3. Choose voice → Alloy, Nova, Onyx, Echo, Fable, Shimmer
4. Generate → AI writes script, generates images, records voiceover
5. Download → Get your finished MP4 video with captions!

API Endpoints

POST /api/generate

Start video generation:

{
  "topic": "5 psychological tricks stores use",
  "style": "documentary",
  "voice": "nova"
}

Returns: { "projectId": "uuid" }

GET /api/status/[id]

Poll generation progress:

{
  "id": "uuid",
  "status": "generating_images",
  "progress": 45,
  "scenes": [...],
  "videoUrl": "/videos/video-123.mp4"
}

Status values: pending → generating_script → generating_images → generating_audio → assembling → complete

Configuration

Voice Options

Voice	Description
alloy	Neutral, balanced
echo	Male, confident
fable	British accent
onyx	Deep, authoritative
nova	Female, warm
shimmer	Soft, gentle

Style Options

Style	Visual Treatment
realistic	Photorealistic, cinematic lighting, 8K
anime	Studio Ghibli inspired, vibrant
horror	Dark, eerie, dramatic shadows
documentary	Realistic, historical accuracy
fantasy	Magical, ethereal lighting
minimalist	Clean, modern design

Deployment

Vercel (Recommended)

npm run build
vercel --prod

Note: FFmpeg isn't available on Vercel by default. For production, consider:

• Serverless FFmpeg layers
• External video APIs (Creatomate, Shotstack)
• Self-hosted on Railway/Render

Docker

FROM node:20-alpine
RUN apk add --no-cache ffmpeg
WORKDIR /app
COPY . .
RUN npm ci && npm run build
CMD ["npm", "start"]

Roadmap

• AI script generation
• AI image generation
• AI voiceover
• FFmpeg video assembly
• Burn-in captions (ASS subtitles)
• Demo mode (mock data)
• Background music library
• Direct TikTok/YouTube upload
• User accounts & history
• Stripe payments
• Custom voice cloning

Costs

Estimated per video (5 scenes, 45 seconds):

• GPT-4o-mini: ~$0.001
• OpenAI TTS: ~$0.02
• Replicate Flux: ~$0.05 × 5 = $0.25
• Total: ~$0.27 per video

License

MIT