ArcReel/ArcReel

AI Agent 驱动的开源可自部署视频工作台：将小说与剧本转为角色、场景、道具资产、分镜、视频和剪映草稿，支持跨镜头一致性、多供应商与费用追踪 | Self-hosted AI video workspace for stories, storyboards and short-form video production

ArcReel – Open‑source, self‑hosted AI video production workbench

What it is – ArcReel is a web‑based platform that lets you turn novels, finished scripts, or product‑related material into short, consistent videos. It strings together content analysis, asset management, storyboard creation, media generation (text, images, video, TTS), cost tracking and final export into a single, auditable pipeline that can be paused and resumed.

Why it matters – Video creation traditionally requires separate tools for writing, illustration, animation, voice‑over and editing. ArcReel unifies these steps and plugs in any AI model that follows the OpenAI‑style API (including MiniMax H3, Sidrune AI, OfoxAI, etc.). This makes it possible to produce narrative or advertising videos with a single, self‑hosted service while keeping full control over data, costs and version history.

Core features

Unified production chain – From raw text to final video, the platform guides you through role/scene/prop extraction, structured script generation, storyboard panels, video clip creation and final rendering.
Agent‑driven automation – Optional AI agents can orchestrate each stage, but users can always intervene, edit, or re‑run a specific step.
Asset consistency & versioning – Assets (character images, scene references) are reusable across scenes; every generation step is version‑controlled and can be rolled back.
Cost & usage monitoring – Central configuration for text, image, video and TTS providers shows estimated and actual token/compute spend before and after generation.
Export to editing tools – Generates a draft for the Chinese “Jianying” (CapCut) editor, allowing further manual fine‑tuning of subtitles, pacing and transitions.
Self‑hosted & extensible – Distributed as Docker images, with SQLite or PostgreSQL back‑ends, and a plug‑in style provider abstraction for adding new AI services.

Quick start (Docker)

# clone and enter the deployment folder
git clone https://github.com/ArcReel/ArcReel.git
cd ArcReel/deploy

# copy example env and start containers
cp .env.example .env
docker compose up -d

Open http://localhost:1241 in a browser.
Default login: admin. If AUTH_PASSWORD is empty, the first start generates a password and writes it back to deploy/.env.
After login, configure your AI providers under Settings → Agent, then create a project and start producing.

Documentation & community

Full docs: https://docs.arc-reel.com (getting‑started guide, workflow details, provider configuration, deployment & ops, architecture, contribution guide).
Video demos and screenshots are included in the README.
Community chat on Feishu (QR code in the repo) for support and sharing creations.

License

AGPL‑3.0 with additional terms (see NOTICE).
Commercial or white‑label deployments that cannot comply with AGPL can be arranged via the project maintainers (support@arc-reel.com).

ArcReel is a genuine AI‑focused open‑source project for automated video generation, suitable for creators, developers, and organizations that want a controllable, self‑hosted alternative to cloud‑only video AI services.