xhongc/ai_story

AI Story – AI‑driven story‑video generation platform

What it is

• A web application that turns a short text prompt (a story theme or outline) into a complete short video. The system automatically writes a script, breaks it into storyboard shots, generates images for each shot, plans camera movements, and finally assembles the frames into a video.

Core workflow

Prompt → Script rewrite → Storyboard generation → Image creation → Camera‑move planning → Video rendering → Finished video

Each step runs as an asynchronous Celery task, and the UI shows real‑time progress. Users can pause, resume, or redo any stage.

Main features

Feature	What it does
Smart script rewrite	Takes a theme or outline and produces a full video‑ready script. Supports several writing styles and lets you pick the underlying LLM (OpenAI, Claude, etc.).
Automatic storyboard	Splits the script into scenes, writes visual descriptions and narration, and creates prompt strings for image generation.
AI image generation	Calls external image‑generation services (Stable Diffusion, DALL‑E, Midjourney, etc.) to produce high‑quality pictures for each storyboard panel.
Camera‑move planning	Generates pan/zoom/tilt instructions for each image so the final video feels dynamic.
Image‑to‑video rendering	Sends the images plus motion plans to video‑generation back‑ends (Runway, Pika, etc.) and produces a video clip.
Project management	Organises multiple video projects, tracks workflow status, allows pausing/retrying, and can export the final video with subtitles.
Prompt‑library	Stores reusable prompt templates with variables (theme, style, length) and tracks their versions.
Model configuration	Centralised UI to add API keys for different AI providers, choose which model to use at each stage, and see usage/cost statistics.

Technical highlights

• Backend – Django application running in a Docker container.
• Task queue – Celery workers (gevent pool) handle long‑running AI calls.
• Cache / messaging – Redis is used for Celery broker and result backend.
• Frontend – Vue.js single‑page app served from its own container.
• Modular pipeline – Each processing step is a plug‑in in a responsibility‑chain pattern, making it easy to add new AI services.
• Load‑balancing & fault tolerance – Supports round‑robin, random, weight‑based, or least‑load routing across multiple model endpoints; automatic retries on failure.
• Deployment – Fully containerised; a single docker‑compose up -d brings up Redis, Django backend, Celery worker, and Vue frontend.

Getting started

1. Install Docker and Docker‑Compose.
2. Create a docker‑compose.yml (the README provides a ready‑made example).
3. Run docker compose up -d.
4. Create an admin user with docker‑compose exec backend python backend/manage.py createsuperuser.
5. Open http://localhost:3000 in a browser to access the UI.

Licensing

• The code is released under CC BY‑NC‑SA 4.0 (non‑commercial use only; commercial use requires a separate license from the author).

Who might use it

• Creators of short‑form videos for TikTok, Douyin, Kuaishou, etc.
• Educators who want to turn textbook text into visual explanations.
• Artists and storytellers looking for rapid video prototypes.
• Parents wanting to generate animated bedtime stories.

Contact / community

• Author’s WeChat: charlesnowed (mention AI Story).
• Telegram channel and group links are listed in the README.
• The project’s GitHub page is https://github.com/xhongc/ai_story.

---

All information above is taken directly from the repository’s README; no additional features have been inferred.