Frame to Movies

Generate a full multi-scene AI movie from a single image and a text prompt.

Frame to Movies is an AI-powered web app that turns a starting frame and a scenario into a complete, concatenated video — scene by scene, preserving visual continuity throughout the entire movie.

How it works

1. Upload a starting frame (image)
2. Write a scenario prompt (up to 1000 chars)
3. Choose the number of scenes and aspect ratio
4. The app automatically:
   • Describes the starting frame (BLIP-2)
   • Generates scene 0 video (VEO 3.1 via Replicate)
   • Extracts the last frame of each scene
   • Uses it as the starting frame for the next scene
   • Concatenates all scenes into one final movie (FFmpeg)

Tech Stack

• Next.js (App Router) + TypeScript
• Replicate API — BLIP-2 (image captioning) + VEO 3.1 (video generation)
• FFmpeg — video concatenation
• Tailwind CSS — dark mode UI

Getting Started

Requirements: Node.js 20+, FFmpeg in PATH, Replicate API token

git clone https://github.com/olidev10/frametomovies.git
cd frametomovies
pnpm install
pnpm dev

Create .env.local:

REPLICATE_API_TOKEN=your_token_here

Open http://localhost:3000

API Routes

• POST /api/generate — accepts scenario, scenes, aspectRatio, image
• POST /api/delete — removes generated movie from /public/generated