Stonewuu/ai-fusion-video

What it solves

Ai-fusion-video (融光) is an Agent-driven video creation platform designed to streamline the entire production pipeline for video creators. It eliminates the fragmentation of tools by integrating scriptwriting, storyboarding, asset management, and AI-powered image and video generation into a single, unified workspace.

How it works

The platform organizes the creative process into a structured hierarchy of projects, episodes, and scenes. It leverages an Agent workspace powered by AgentScope, which allows creators to interact with AI via streaming conversations and multimodal inputs (images, videos, audio, and files).

Key technical components include:

• Agentic Workflow: Supports skills, MCP (Model Context Protocol), and sub-agents to handle complex, multi-step creative tasks.
• Model Integration: Connects to various LLMs (OpenAI, Anthropic, Gemini, DashScope, Ollama) and specialized image/video generation services via API.
• Asset Pipeline: Manages media assets using local storage or S3-compatible object storage, with FFmpeg and FFprobe for video processing.
• Full-stack Architecture: Built with a Java 21/Spring Boot backend and a Next.js 16 frontend.

Who it’s for

It is primarily for video creators and production teams who want to use AI to accelerate their workflow from initial script to final visual assets.

Highlights

• Unified Workspace: Combines scriptwriting, storyboarding, and asset generation in one place.
• Agent-Driven: Features a sophisticated Agent workspace with multimodal context and tool-calling capabilities.
• Flexible Model Support: Supports a wide range of text, image, and video models without requiring hard-coded keys.
• Enterprise-Ready Deployment: Provides Docker Compose configurations for easy setup, including support for separate frontend and backend deployments.