Captured on 19 October 2026, the screenshots show the actual editor with an original generated STUDIO SESSION demo and procedural audio. They demonstrate the interface, not AI editing quality.
Contents
1.3 Beta update
The repaired1.3 beta gives added videos their detected source sound by default and exposes sound, mute and gain controls. Existing silent overlays keep their saved choice. The concise AI draft notice expands in the inspector, leaving the preview clear. Exact Mac ARM64/Intel automation and configured-Windows synthetic exports passed; clean-OS Windows installation and human acceptance remain unqualified.
Highlights
- Six focused workspaces. Media, Edit, Layout, Audio, Captions and Output share one player and one timeline.
- Independent A/B tracks. Cut, trim and move a camera and a screen recording together or separately, with Undo/Redo and track locks. Timeline controls
- Camera and screen layouts. Stacked, side by side, picture in picture, camera only or screen only, for a selected range or the entire edit.
- Track mixer. Balance Original, Music, Voiceover, Effects and Master with live level meters. Mute applies to preview and export; Solo is for preview only.
- Text and captions. Add titles and captions, import SRT or VTT, and correct and style an optional AI transcript.
- Green screen and stabilization. Apply a chroma key to any video source, or create an optional stabilized local copy. Green screen · Stabilization
- Local MP4 export. Landscape, vertical, square, portrait or the source shape, from 720p to 4K at up to 60 FPS, with no CUTROOM watermark.
- Gameplay vision for Shorts. A compatible installed local Story model looks at sampled frames to propose action moments and filter menus and loading screens. Frames stay on your computer. Models
- AI is optional. Manual editing needs no AI model, API key or account.
Quick start
Download for Windows or macOS
- Extract the entire ZIP into a fresh folder. Keep App, START HERE.html and the launcher together.
- Open your platform's launcher beside App and let setup finish.
- Edit in your browser. Keep the launch window open while you work; closing the browser alone does not stop CUTROOM.
These are source-based beta packages. First setup needs internet to install dependencies. Python, FFmpeg, the local Story engine and AI models are not bundled. If the browser does not open, copy the address shown in the launch window once setup finishes. The Mac launcher is unsigned and is not a signed or notarized .app; the platform guide explains the security prompts.
[!IMPORTANT]
Before the first Windows launch, make sure WinGet is available and install the official Microsoft Visual C++ x64 runtime. The CUTROOM launcher does not install this prerequisite. See the platform guide for setup and troubleshooting.
Release notes. The 18 October beta remains available for rollback. To update, extract a fresh copy separately and back up your projects before moving them.
Mac uses the exact tested combined ZIP: extract it, open mac, then run START CUTROOM.command beside App. This is a feedback beta; see the release notes for exact source/archive checks and the incomplete Windows clean-OS/prerequisite qualification. No models were downloaded for this repair.
The release includes SHA-256 sidecars for the Windows ZIP, Mac ZIP and combined ZIP. The dated release notes identify the exact app source and recorded checks.
An optional Windows + Mac ZIP contains both platforms. Open windows or mac inside it before launching.
Run from source
- Windows: run
run_windows.bat from the repository root.
- macOS: follow the Mac setup guide.
- Linux: install Python 3.12 with
venv and pip, FFmpeg/FFprobe and Bash, then run:
chmod +x setup_linux.sh run_linux.sh
bash setup_linux.sh
bash run_linux.sh
Setup creates a local virtual environment and installs the Python dependencies. FFmpeg needs the encoding and subtitle filters listed in the Linux prerequisites.
How it works
- Bring your footage. Start with one recording, or add a camera and a screen recording as sources A and B.
- Build your edit. Edit by hand, or review an AI-assisted YouTube cleanup or Short/Reel draft.
- Make it yours. Trim and arrange clips, adjust framing and sound, add captions, then export an MP4 and watch the result.
The project library stays on the left, the preview in the centre, properties on the right and the timeline below. Click footage or a caption in the timeline to open its tools. Select imported audio to adjust that clip's volume and fades; the Audio workspace adjusts whole mixer groups. Drag a divider, or focus it and use the arrow keys, to resize the preview, timeline and properties. Compact windows use a scrolling layout.
| Workspace | What you do there |
|---|
| Media | Import recordings, images, music and voiceover. Search and filter the project library. Your original files stay untouched. |
| Edit | Split, trim, move and restore clips. Protect a track with a lock. Apply a green screen in Effects. |
| Layout | Choose the composition, assign camera and screen roles, set the framing and adjust the sync offset. |
| Audio | Mix Original, Music, Voiceover, Effects and Master with live meters, Mute and Solo. |
| Captions | Add text and captions, import SRT or VTT, correct the transcript and style the subtitles. |
| Output | Choose the shape, resolution, frame rate and quality, then export. |
User guide · Keyboard shortcuts · Shortcut profiles
AI, your way
| Mode | What it does | What it needs |
|---|
| Manual editing | You make the cuts, add media and captions, and export. | No AI model, API key or account. |
| Local AI | Transcribes speech and proposes an edit using models on your computer. | Separate model downloads, disk space and processing time. Story AI also needs a separate local engine; CUTROOM links to its installer. |
| Cloud AI (optional) | Uses Groq or a compatible provider for transcription and edit planning. | Your own API account, compatible models and your explicit consent. Provider limits and billing apply. |
- You stay in control. A new AI draft does not replace your manual timeline until you choose Apply new AI draft, and Undo stays available.
- Plan for model downloads. In Auto mode, CUTROOM uses the stronger transcription and Story models only when they are already installed and your graphics card has room for them. It suggests a larger Story model when it would fit. Models are separate downloads; a missing selected speech model can download on its first use. Download behavior
- Know your limits. Mac transcription currently runs on the CPU. A chat subscription does not include API access, and free tiers have usage limits.
AI connections and privacy · Models and download sizes
Privacy
- Local work. Editing, preview and MP4 export run on your computer. Projects and working media are stored locally.
- Cloud AI only with consent. When cloud AI is enabled, selected audio is sent to your provider for transcription, and transcript text and editing context for planning. Video frames are not sent. The provider's retention and privacy terms apply.
- No CUTROOM account. There is no CUTROOM subscription, account or export watermark.
Requirements
| Platform | Status |
|---|
| Windows | Source-based beta ZIP with launcher. WinGet and the Microsoft Visual C++ x64 runtime are prerequisites; see Quick start. Physical-device coverage remains pending. |
| macOS | Source-based Mac beta ZIP for macOS 15 or later on Apple Silicon and Intel. Finder, Gatekeeper and Safari behavior still need manual verification; see the dated release for automated checks. |
| Linux | Source checkout only: Python 3.12, FFmpeg/FFprobe and Bash. No Linux download, and the full Linux installation is not yet verified end to end. |
See the platform guide for exact prerequisites, commands and troubleshooting.
Beta status
CUTROOM 1.3 Beta is beta software. Start with a short recording and keep your originals. Review transcripts, AI selections, source roles, synchronization and the exported MP4 before you publish. AI proposes a draft for you to edit; the default YouTube cleanup keeps your recording's order.
With two recordings, confirm that the sources show the same moment before relying on automatic synchronization. A camera appears only where its track covers the edit; an empty B track does not create camera footage. Detection and controlled export checks do not establish the alignment of your recordings.
The editor has two main video sources plus media and text layers. Added media and custom text layers stay within the existing edit and do not automatically follow source reordering. Saved transcript-caption corrections follow the surviving source speech across trimming and reordering. Automated checks and synthetic renders do not prove real-footage quality, installation on every computer or AI accuracy.
Exact package checks and source identities belong to each dated release. They do not establish installation on every computer or human acceptance of AI, captions and sound. Physical-device setup, Mac security prompts and representative English/Hebrew editing and listening remain part of the manual acceptance protocol.
Beta status · Known limits · Mac validation · Changelog
Documentation
Contributing and support
License and credits
CUTROOM is released under the MIT License. Third-party tools and AI models keep their own licenses; see the third-party notices.
Thanks to the open-source projects behind CUTROOM, including FFmpeg, faster-whisper, ivrit.ai and the local Story engine listed in the third-party notices. CUTROOM is independent of external AI providers.
