The open source Descript alternative.

Drop your clips into LogCut and it transcribes them. Cut the transcript, and the timeline cuts with it. When the story works, switch to the timeline for the finishing pass — trims, tracks, captions, export. AI sits alongside the whole way: it can rough out a first cut, strip silences, or find the take you half-remember, and every change it makes lands as a normal edit you can inspect, adjust, or undo.

And because LogCut is open source, the editor’s code is public. You can read it, file issues against it, build on it — and trust that the tool you learn today can’t be discontinued, paywalled, or “sunset” out from under you.

Most of editing talking-head video, tutorials, podcasts, and interviews is deciding what gets said and in what order. That work is faster in a document than on a timeline. In LogCut, the transcript is the edit: delete a sentence and the video cuts; reorder paragraphs and the story reorders; select a rambling take and replace it with the tight one. Filler words and dead air are highlighted so clearing them is one pass, not an afternoon of razor cuts.

You never lose the frame-level view — text-based editing and the timeline are two views of the same cut, always in sync. Write the story first, then polish the pictures.

Text-based editors tend to treat the timeline as an afterthought. LogCut doesn’t. Under the transcript is a proper multi-track timeline with the behavior editors expect: precise trims, snapping, keyboard-driven navigation, and direct control of every clip boundary the text view created. B-roll, music, and titles live on their own tracks. When a cut needs a two-frame nudge instead of a sentence-level decision, you make it exactly where you’d expect to.

The goal is simple: the speed of editing a document, without hitting a ceiling the moment your project needs real precision.

Transcription is the foundation, so it’s built in, not bolted on. LogCut transcribes in multiple languages, with control over how lines are segmented — including a short-sentence mode that breaks speech into caption-sized lines instead of paragraph slabs, so subtitles are readable the moment they’re generated.

LogCut’s AI works like an assistant editor, not an autopilot. Tell it what you want — “cut the silences,” “assemble the answers about pricing,” “tighten this section to ninety seconds” — and it proposes an edit you can review. Its changes are ordinary cuts on the timeline: visible, adjustable, reversible. Nothing is hidden inside a black box render.

Where AI genuinely helps — first assemblies, silence and filler removal, finding moments, drafting captions — it does the boring work. Where taste matters, you keep the keyboard.

LogCut is local-first: your media lives on your disk, and editing doesn’t require uploading an hour of source footage before you can make a cut. Features that need heavy AI models can use hosted processing — those are opt-in and clearly marked, never the default. For client work under NDA, footage that can’t leave the building, or just a slow connection, this is the difference between a tool you can use and one you can’t.

The code is public at github.com/logcut. Development happens in the open: real issues, a visible roadmap, and decisions you can argue with. Open source isn’t a marketing word here — it’s the reason a tool you invest hours learning stays yours.

Descript is a polished product, and if it fits your workflow you should keep using it. LogCut exists for the people it doesn’t fit: editors who want text-based editing and a serious timeline in one tool, teams who can’t upload client footage to someone else’s cloud, and anyone tired of per-seat subscriptions and AI credit meters on software they open every day.

Same core idea — edit media like a document. Different contract: the code is open, files stay local, and the project is built in public.