01
Your footage stays on your machine
Cutroom is a video editor that runs entirely in your browser. When you import a clip, it is copied into this browser's own storage (IndexedDB and the origin private file system) and read from there. Cutting, previewing and exporting all happen on your device. No frame is uploaded to edit it.
Projects, media and settings live in this browser on this machine. They do not sync to other devices, and clearing this site's data in your browser deletes them for good. Export anything you want to keep.
02
No accounts, no analytics, no tracking
There is nothing to sign up for. Cutroom loads no analytics, no advertising tags, no bot checks and no third-party scripts. The server that delivers the app keeps the standard short-lived request logs any web server writes, and nothing about your projects.
The only cookies or local storage keys Cutroom sets are its own working state: your projects, panel sizes, shortcuts and, if you connect one, your Deference key.
A few optional assets come from public services when you ask for them: picking a Google font for a text layer loads it from Google Fonts, and the platform safe-area guides show logos from Brandfetch. Neither request carries your media.
03
AI effects send only what they need
The Studio tab offers AI effects such as captions and translation. They never run on their own: each one starts only when you press its button, after showing an estimate of its cost.
When you run one, Cutroom sends only the material that effect needs, for example the audio of the clip you are captioning or the caption text you are translating, through this app's server to the Deference gateway, which passes it to the model provider that does the work. The result comes back into your project. Your footage as a whole is never sent.
Effects are paid per use in INFERENCE from your own Deference account. Your Deference key is kept in this browser's local storage and sent with each run so the gateway can charge it. Cutroom's server forwards it and does not store it. Disconnect the key in the Studio tab to remove it from this browser.
What the gateway and the model providers do with a request is governed by their own terms. Deference publishes its policies at deference.si.
After an export you can choose to stamp it: your wallet records the file's SHA-256 fingerprint and the number of AI runs it used on the Sepolia test network, where anyone can read them. The video itself is never sent. Skip the stamp and nothing is recorded.
04
Effects that run locally
Some tools run on your device instead: local caption transcription downloads an open speech model once and runs it in your browser, and silence and filler cutting works on transcripts you already have. Neither sends your media anywhere.
05
Changes and questions
If this page changes, the new version will be published here with the date below. Questions about privacy go to the Deference Works team through deference.si.
Last updated October 2026