Ai Video Studio

ai-video-studio

[中文](README_CN.md) | English

AI video production workflow platform for professional short drama teams. Timeline-first, harness-tested, agent-assisted.

ai-video-studio is not a video generation model. It is an engineering workflow system for AI short-drama, film, and TV production. It brings story, script, audio, timeline, storyboard, assets, rendering, export, and test evidence into one production chain.

The core idea is to help creative teams collaborate around a playable `Timeline`, not around one-off model outputs. The current main chain is `audio -> timeline -> clip -> render -> export`: `Timeline` is the single source of truth (SSOT) for playable episode output, while Storyboard remains a visual support view and compatibility surface.

What It Does

• Manages production objects: virtual IP, stories, episodes, scripts, audio, Timeline, storyboard, and media assets.
• Turns AI outputs into traceable production assets instead of temporary prompts and files.
• Organizes clip editing, asset replacement, rendering, export, and quality evidence around Timeline.
• Supports reproducible engineering collaboration through harnesses, logs, browser evidence, and agent ledgers.

Repository Layout

• ai-pic-backend/: FastAPI + SQLAlchemy + Alembic + Celery (MySQL/Redis)
• ai-pic-frontend/: Next.js 16 App Router + TypeScript + Tailwind
• docker/: local development and production Docker stacks with Nginx entrypoint
• docs/: design, API, and testing documentation index
• tasks.md: canonical product task board

Quick Start: Lite Mode In 5-10 Minutes

Lite mode is the fastest local path. It uses SQLite, runs Celery tasks eagerly in-process, and enables AI mock fallback by default. It does not require MySQL, Redis, or a separate worker.

1. cd docker
2. ./init_env.sh lite
3. ./dev_lite_in_docker.sh

Open:

• Web app through Nginx: http://localhost:8089
• Backend API: http://localhost:8000
• Swagger: `http://localhost:8000/d