About
Supercut is a video editor you talk to. You drop in a clip, type the edit you want, and get the cut back. The part that makes it unusual is what does not happen: your video is never uploaded. Every frame is decoded, edited and exported on the machine you are sitting at, which is why it keeps working with the network unplugged.
This website is operated by Harsh Manwani, an independent developer in India. Supercut is not a ByteDance product, it is not affiliated with CapCut, and there is no venture money behind it. That matters for a practical reason rather than a romantic one: a small paid product has no reason to build the upload pipeline that a large free one needs, and no second business that your footage would be useful to.
Support is answered by the person who wrote the code. Write to support@supercut.cc and that is who replies, usually within one business day. There is more on how to reach us, including press and security, on the contact page.
Supercut sells a software subscription. It is a video editor that runs on your own computer, in a web browser at supercut.cc/app or as a desktop app for macOS and Windows. You describe an edit in plain English and the software performs it: trimming, cropping, captions, compression, format conversion, colour and audio work, and screen recording. Nothing is shipped, because there is no physical product to ship.
Supercut does not support adult content. The Service must not be used to create, edit or export pornographic, sexually explicit or nudity-related material, and no adult video editing service is offered on this website or in the app. This applies to the free tier and to every paid plan.
There is a free tier that needs no account, and three paid plans, priced on the pricing page. Reach us at supercut.cc/contact.
Every decode, edit and export runs in your browser through FFmpeg compiled to WebAssembly, or through a native binary in the desktop app. There is no upload step, no render queue and no bucket with your files in it. The one network call an edit makes carries your prompt text and clip metadata, never a frame.
When you type "crop this vertical and burn in captions", the model receives that sentence plus the duration, dimensions and filename of your clip. It returns a structured operation. A deterministic mapper turns that operation into FFmpeg arguments. The pixels are never part of the conversation.
The 43 editing operations are ordinary FFmpeg work, not model output. That is why the same prompt on the same clip gives the same result twice, and why a bug in an edit is a bug we can fix rather than a phrasing to retry.
Supercut is funded by people buying it. There is no ad network, no data broker and nothing to sell, which is a large part of why the privacy claim is cheap for us to keep.
If you have seen this product under another name, you saw the same product. The engine never forked.
It started as a single question: could you edit a video by describing the edit, without uploading it anywhere? The first version answered that, and capcutgpt.com is still live running the same engine.
The name had to change. CapCut is ByteDance’s product and this has never been affiliated with it, so a name that borrowed theirs was misleading whatever the intent. UnoCut was the rebrand, along with a command-first redesign of the editor.
The current name, and the one the product ships under: a web app at supercut.cc, a Mac App Store release, and notarized desktop builds for macOS and Windows that add a screen recorder and a local MCP server.
Every page of this site is published as plain text so you do not have to parse the markup: llms.txt for the index, llms-full.txt for the whole corpus, the Markdown catalog for one page at a time, and ai.txt for what you may do with it. There is a hosted MCP server at supercut.cc/mcp and a full index of developer resources at supercut.cc/docs.
See also the privacy policy, the terms, and support.