A complete video editor that runs in your browser: a multi-track timeline with clips, text, stickers and music; effects, masks and keyframe animation; automatic captions from a speech model running on your device; and MP4 (H.264) or WebM export encoded right here with WebCodecs. There is no account, no watermark and no upload - your footage never leaves this browser, and projects save themselves locally so you can close the tab and pick up later.
This is the job the single-purpose pages cannot do in one place: our trimmer, splitter and merger each handle one file and one operation, while the editor holds a whole project - several clips, overlays, an audio bed and captions - and exports one finished video. For a quick single cut the small tools are still faster.
It also has something the desktop apps do not: an AI copilot running on YOUR device (WebGPU). Type "cut the first three seconds and add the title Hello at 0:02" and a local language model turns it into ordinary, undoable edits - it never sees your video and its instructions are schema-checked before anything is applied.
What is inside the editor
Timeline: a main video track plus overlay and audio tracks; trim, split, reorder, snap, ripple-edit, group moves and a full undo history. Text and captions: styled titles with fonts, backgrounds and animation, plus automatic subtitles - a Whisper speech model downloads to your browser once and transcribes locally in several sizes up to large-v3-turbo on WebGPU. Look: effects, masks, speed changes with pitch preservation, and keyframes on position, scale and rotation. Sound: volume, mute, audio separation, and a sound-effects library searched through our server proxy (the one feature that touches a server, because the catalogue needs an API key - your media still never leaves the browser). Export: MP4 or WebM in four quality levels, encoded by your own machine.
Where your work lives
Projects and imported media are stored by the browser itself, in IndexedDB and the Origin Private File System, on your disk. Clearing site data deletes them; nothing has a copy anywhere else. The editor is a static build served from our asset host - there is no server-side session, which is why it keeps working offline once loaded.
Credit and licence
The editor is OpenCut (MIT), the open-source CapCut alternative, self-hosted at a pinned version with its analytics removed and an on-device AI copilot added. The full patch we apply is published alongside the build.
Video Editor
Edit a whole video project in your browser: several clips on a multi-track timeline, titles and stickers, background music, automatic captions, effects and keyframes - then export MP4 or WebM encoded on your own machine. No account, no watermark, no upload: footage, projects and exports all stay on your device, saved by the browser itself so your project is still here tomorrow.
A real editor, not another single-purpose tool
The trimmer, splitter and merger on this site each do one operation to one file. The editor is for the composite job: assemble a cut from pieces, put a title over the intro, lay music under it, caption the speech, and ship one finished video. Trimming, splitting at the playhead, reordering, ripple edits, speed changes that keep pitch, masks, effect keyframes and a full undo history are all in the timeline.
The AI that lives on your device
Two models can run inside the tab, both optional and both local. A Whisper speech model writes captions from the audio - it downloads once, then transcribes on your machine in your choice of size, up to large-v3-turbo when WebGPU is available. And an AI copilot we added takes typed instructions - "cut the first 3 seconds, add the title Hello at 2s, mute everything" - and turns them into ordinary timeline edits: the language model runs on WebGPU locally, its output is checked against a strict schema before anything is applied, and every change it makes is undoable with Ctrl+Z. On machines without WebGPU the copilot button explains itself and every manual feature keeps working.
The engine, and credit
This is OpenCut (MIT licence), the open-source CapCut alternative, self-hosted by us at a pinned version as a fully static build: we removed its third-party analytics and its account backend - the editing engine never needed either - and added the copilot. The one server call left is the sound-effects catalogue search, which goes through our own proxy because the catalogue needs a server-side key; your media is never part of that request. The complete patch is published next to the build for anyone to read.
Frequently Asked Questions
Is my video uploaded anywhere?
No. Importing a file hands it to the browser, editing happens in the tab, the export is encoded by your own machine with WebCodecs, and projects are stored by the browser on your disk. The only server call in the whole editor is the optional sound-effects catalogue search - and that request contains your search words, never your media.
Where are my projects saved, and can I lose them?
In your browser's IndexedDB and Origin Private File System, on this device. They survive closing the tab and rebooting, and the editor works offline once loaded. They do NOT follow you to another browser or computer, and clearing this site's browsing data deletes them - export anything you cannot afford to lose.
How is this different from the trimmer, splitter and merger tools?
Those pages do one operation to one file and are faster for a single cut. The editor holds a project: several clips on a timeline, text and stickers on overlay tracks, music underneath, captions, effects and keyframes - and exports one finished video. Use whichever matches the job.
What does the AI copilot do, and what does it need?
You type an instruction like "cut the first 3 seconds and add the title Hello at 0:02"; a language model running on your device turns it into normal timeline edits. Its output must pass a strict schema before anything is applied, everything it does is undoable with Ctrl+Z, and it never sees your footage - only a text description of what sits on the timeline. It needs WebGPU (a recent Chrome or Edge on a computer) and a one-time model download of roughly 700 MB; without WebGPU the button is disabled with an explanation and all manual editing keeps working.
How do automatic captions work?
A Whisper speech model downloads to the browser once and transcribes the audio locally - pick a size from tiny to large-v3-turbo, the bigger ones wanting WebGPU. The audio never leaves the device, and the captions land on the timeline as styled, editable text.
What formats can I export, and is there a watermark?
MP4 (H.264) for compatibility or WebM (VP9) for smaller files, in four quality levels, with or without audio. No watermark, no duration cap from us - the practical limits are your machine's speed and the browser's storage.
Why does it want a computer rather than a phone?
A timeline editor needs screen space and the phone browsers cut back the APIs it relies on (WebCodecs encode, OPFS, WebGPU). On a phone the editor shows a message instead of a broken interface; the single-purpose video tools on this site work fine on phones.
What exactly is OpenCut and what did you change?
OpenCut is the open-source (MIT) CapCut alternative. We self-host its editor as a static build at a pinned version and changed four things: removed its analytics scripts, removed its account backend (the editor never needed one), routed the sound-effects catalogue through our own key-holding proxy, and added the on-device AI copilot. The complete patch is published next to the build.