Nomi: Open-Source, Local-First AI Video Creation Workbench

Nomi is an open-source, local-first AI video creation tool that integrates scriptwriting, image generation, video generation, and editing into a unified workflow. It addresses the fragmentation pain point of existing AI video creation tools and provides creators with a continuous canvas environment from idea to final video. Key features include natural data flow between creation stages, local-first data sovereignty, multi-vendor model support, and AI Agent-assisted collaboration.

With the development of generative AI technology, there are more tools for video creators, but existing workflows are fragmented: scriptwriting, image generation, video generation, and editing require switching between different tools, reducing efficiency and easily losing creativity. Nomi was born to solve this problem.

Nomi's core design breaks down data barriers between creation stages, enabling natural flow between text, images, videos, and timelines: script prompts directly generate image/video nodes, generated content can be seamlessly dragged into the timeline for editing, and edited clips can flow back as reference materials. The local-first architecture keeps user data (scripts, materials, projects, etc.) local, bringing advantages like data privacy, offline work, cost control, and customizability (open-source code), while also supporting integration with external model APIs.

Nomi is not tied to a single model platform; it supports integration with various image/video generation services (private gateways, enterprise APIs), and simplifies the integration of new vendors via Nomi Agent. The built-in AI Agent assists with script breakdown, node suggestions, prompt optimization, etc., but the final decision-making power rests with the creator. Target users include AI short video creators, manga/novel adapters, those needing workflow integration, data-sensitive teams, and private model users.

Nomi uses a modern web tech stack: apps/web (creator interface), apps/hono-api (local API service), apps/agents-cli (Agent bridge), packages (shared schema). Environment requirements: Node.js 20+, pnpm 10+, PostgreSQL 16+. Typical workflow: 1. Write scripts/storyboards in the creation area; 2. Agent suggests image/video nodes; 3. Generate media; 4. Drag into timeline for editing; 5. Export final video—no tool switching needed throughout the process.

Nomi uses the Apache-2.0 open-source license and welcomes community contributions (bug fixes, feature additions, vendor adaptations). The project is in the early stage but has a clear architecture. Nomi represents an important direction for AI video creation tools: moving from scattered cloud services to unified local workflows, and from black-box automation to human-machine collaboration. It provides a valuable reference for improving the experience of AI creation tools.