Open-source project / Lumen

Released for Windows · MIT licensed

Lumen: an open-source AI video editor.

We built Lumen so you can ask an AI to work on your video, inspect its edits and keep cutting by hand. The Windows app combines a multi-track editor with a Copilot that can examine frames, listen to the mix and change the timeline.

Discuss a similar project

Get Lumen for Windows View source on GitHub

Windows 10 / 11 · 64-bit · MIT license. Download details and release notes are on GitHub.

View all 5 screenshots

Lumen video editor with a multi-track timeline, 3D title preview and AI Copilot reviewing an edit — The Lumen editor with its launch-demo project: media library, rendered preview, Copilot and layered timeline in one workspace. Open full-size image in a new tab

PROJECT SCREENS

More of Lumen.

Screenshots from Lumen's public launch demo. Open any image at full size to inspect the interface; the current release may include newer controls.

Lumen Copilot showing project inspection, a contact sheet, timeline edits and a rendered frame check — Review the AI's work
The Copilot shows the tools it used and the frames it checked. The Undo all control lets the user revert the response's edits.
Open full-size image ↗

Lumen motion graphics panel with lower-third, title card, kinetic type and counter templates — Build motion into the edit
The Generate panel offers motion templates with a preview inside the editing workspace.
Open full-size image ↗

Lumen integrations screen showing Blender setup alongside AI models, HyperFrames and MCP connections — Connect the creative tools
The integrations hub connects Blender rendering and other creative tools to the project.
Open full-size image ↗

Lumen export dialog with video format, codec, resolution, frame rate and hardware encoding controls — Take the finished video with you
Export controls keep the output format, quality and estimated file size visible before rendering.
Open full-size image ↗

THE BRIEF

The work behind Lumen.

Video editing involves decisions across picture, sound and timing. We wanted an assistant that could inspect the current edit and act on it, with changes you can review and undo inside the editor.

Lumen is an open-source project released by Async. The public repository includes the source code, architecture notes and Windows releases. Developers can inspect how the editor works or contribute to it under the MIT license.

PRODUCT SCOPE

What we built.

An editor you can finish in

A magnetic timeline supports layered video, titles and audio. Trim clips, adjust keyframes and colour, then export with frame-exact rendering and hardware encoding where supported.

AI with editing tools

The Copilot can inspect frames and transcripts, check audio and make timeline edits. Those edits use the same command system as manual changes, with an undo history you can review.

Captions and motion graphics

On-device Whisper transcription supports timed captions. Blender integration adds 3D titles and scenes, while HyperFrames renders motion graphics such as lower thirds and kinetic type.

Your choice of AI

Connect a supported AI provider or local model. Lumen also exposes editing tools through MCP, so an external agent can work with the project. Cloud providers process the data sent to them and may charge for usage.

WORKFLOW

A connected sequence of tasks.

01Bring in footage
Import media and arrange a timeline.
02Ask for an edit
Give the Copilot a task, such as tightening pauses or adding captions.
03Review the changes
Inspect the result, undo a step or refine it by hand.
04Export
Choose the output format and render the finished edit.

DESIGN & ENGINEERING

Decisions that shaped the build.

Give people and AI the same editing commands.

A shared command system handles timeline changes from the interface and from agents. AI edits stay part of the project's history, so the user can inspect or undo them.

Let the Copilot check rendered frames.

The AI's frame-inspection tools use the export compositor. This lets it review titles, effects and transitions together rather than judging a source clip in isolation.

Keep local editing independent of cloud AI.

The desktop app references media on disk and can transcribe with an on-device model. Users choose whether to connect cloud services for AI features; those services have their own data handling and usage costs.

RELATED SERVICES

AI agents & automation Custom software

ElectronReactTypeScriptWebCodecsWhisperMCP

YOUR PROJECT

Tell us what you need to build.

Share the users, the current workflow and the part you want to improve. We can discuss a useful first release.

Discuss your project