AI Clip Assembler is an open-source local-first desktop video editor project for working through long recordings such as drone, travel, action, or event footage. The goal is simple: make the first pass through messy raw material less manual by helping surface usable clips, scores, and cut suggestions before a polished final edit exists.
Built as a foundation, not presented as a finished app
I am the creator and maintainer of the concept, architecture, and codebase foundation. The work completed so far is the part that makes later editor features possible: the Electron and React frontend structure, a FastAPI backend skeleton, the core architecture and harness layer built around a standardized JSON contract, local Qwen setup, and the first version of the brand assets.
That matters because this project is intentionally not framed as "AI magic edits your video for you." The real problem is building a workflow where automated analysis can help, but the system stays inspectable, local-first, and useful even before every advanced editor interaction is finished.
Important product and engineering choices
A core decision was to standardize how analysis modules and model-backed harnesses communicate. The JSON contract keeps the project from being tied to one provider or one model runtime too early, which is important when local setups vary and the product still needs a deterministic path for testing and iteration.
The second big choice was to keep the stack local-first. Using Electron on the desktop and local model/runtime setup means the product can be developed around real media workflows instead of assuming cloud processing is always the right answer.
Current status
This is active work in progress, not a finished released editor. Runtime integration, the first packaged macOS build and DMG flow, the motion-analysis module, richer timeline interactions, and export generation are still current work. The public repo and demo show where the product is going, but the editor is still at the foundation stage rather than a complete end-to-end pipeline.
