What you need to run Cutroom
Cutroom runs on Windows 10 or 11, 64-bit. The editor and the transcription run on the CPU, so you do not need a graphics card to use it. It ships with its engines, including a multilingual transcription model, so a fresh machine works offline the moment it is installed. A GPU is a speed-up, not a requirement: it makes exports faster and powers AI upscaling. To drive Cutroom with an AI agent you also need Node.js. Here is the honest minimum and what makes it comfortable.
- OS
- Windows 10 or 11, 64-bit
- CPU
- Any modern 64-bit CPU
- GPU
- Not required
- Disk
- A couple of GB free
- GPU
- A Vulkan-capable GPU
- Memory
- 16 GB
- For AI
- Node.js
The short version
The floor, and what makes it comfortable
The minimum is genuinely modest, because the editor and transcription run on the CPU. Memory is the one thing that scales the transcription model: at 8 GB Cutroom runs the high-accuracy multilingual model, and on a lighter machine it automatically steps down to a smaller model instead of running out of memory, so it never crashes on a weak PC. If you have a graphics card, you get faster exports through hardware encoding, and AI upscaling turns on, which needs a Vulkan-capable GPU. More memory also helps with 4K footage and many tracks at once.
- Minimum: Windows 10/11 64-bit, a modern CPU, 4 GB RAM (the model scales down to fit), room on disk
- Recommended: 8 GB RAM for the high-accuracy multilingual model, 16 GB for 4K and many tracks
- A GPU for fast exports and AI upscaling; Node.js only to drive it with an agent
Minimum vs recommended
No hard memory gate.
The only thing Cutroom checks before it acts is free disk space, before it downloads an engine. It does not block on memory, so the recommended specs are guidance, not a gate.
Everything ships in the app
Cutroom bundles its engines inside the installer, so a fresh machine transcribes, captions, and cuts offline the moment it is installed, with no first-run download. That includes a multilingual transcription model, so Arabic and more than 90 languages work out of the box. The download is larger because of it, roughly 2 GB, and the app takes a similar amount on disk. Budget a couple of extra gigabytes for your own projects and exports. An even larger, top-accuracy transcription model and the AI upscaler are optional, and only download if you ask for them.
- The installer is about 2 GB because the engines and multilingual models ship inside it
- A fresh machine works offline immediately, with no first-run download
- An even larger transcription model (about 3 GB) and the AI upscaler are optional
What ships in the app
- Installer download
- ≈ 2 GB
Questions about requirements
Windows 10 or 11, 64-bit, a modern 64-bit CPU, about 4 GB of RAM (8 GB recommended), and roughly 2 GB of free disk space for the app plus room for your projects. The editor and transcription run on the CPU, so no graphics card is required, and the engines ship inside the installer, so it works offline from the first launch.
No. A portable, multilingual transcription engine ships in the app and the editor runs on the CPU, so Cutroom works on machines without a dedicated GPU. A graphics card is a speed-up: it makes exports faster through hardware encoding and is needed for AI upscaling, which uses a Vulkan-capable GPU.
Not today. Cutroom is a Windows desktop app for Windows 10 and 11, 64-bit. There is no Mac or Linux build.
About 4 GB is the practical floor and 8 GB is recommended. Memory is what decides the transcription model: at 8 GB Cutroom runs the high-accuracy multilingual model, and on less it automatically steps down to a smaller model so it never runs out of memory. 16 GB is comfortable for 4K footage and many tracks at once.
Yes. The engines, including a multilingual transcription model that handles Arabic and 90+ languages, ship inside the installer and are set up on first launch, so editing, transcription, captions, and cutout all work with no connection. You only need the network to connect an AI agent, run a generation, or install an even larger optional model.
The installer is about 2 GB because it bundles the engines and the multilingual transcription models so everything works offline. The app takes a similar amount on disk. Budget a couple of extra gigabytes for your own media and exports. An even larger, top-accuracy transcription model (about 3 GB) and the AI upscaler only download if you ask for them.
Only to drive Cutroom with an AI agent over MCP, since the connector runs on Node. Editing by hand needs nothing extra. Cutroom checks for Node and points you to it if it is missing.