aqm857886159/Nomi
Open-source AI video workbench. Bring any model or your local ComfyUI, and let Claude Code / Codex / Cursor direct it over MCP — storyboard, references, generation, editable first cut on a real timeline. Local-first: projects, prompts, and keys stay on your machine. No account, no telemetry.
What it solves
Nomi reduces the cost and complexity of AI video production by providing a local-first desktop workbench. It eliminates the need for expensive, locked-in cloud platforms by allowing users to mix and match various generation sources—including free credits, cheap APIs, and local ComfyUI setups—within a single pipeline for scripting, storyboarding, and editing.
How it works
Nomi acts as an orchestration layer on a user's machine. It connects to multiple AI providers (such as Runway, fal, Replicate, and ModelScope) or local instances of ComfyUI. Users can create a "draft $\rightarrow$ reference $\rightarrow$ finish" workflow, using low-cost models for initial sketches and high-quality models only for final renders. Additionally, it supports the Model Context Protocol (MCP), allowing AI agents like Claude Code or Cursor to drive the generation and editing process via 25 built-in tools.
Who it’s for
It is designed for AI video creators and companies that need a scalable, customizable, and private way to produce AI video without being tied to a single provider's pricing or telemetry.
Highlights
- Local-First Architecture: All footage, workflows, and provider configurations are stored locally on the user's machine.
- Hybrid Generation: Freely combine local ComfyUI workflows with various external API providers to optimize costs.
- Agent-Driven: Integrates with AI agents via MCP, enabling external LLMs to orchestrate the video production pipeline.
- Open-Source & Customizable: The codebase is open, allowing teams to build internal AI-video platforms or add custom skills.
Related
- Project
- Project
- Project
- Project
- Project