Transcribe, cut and render on your Mac
AgenticCutter transcribes, cuts and renders on your Mac without uploading the video file for these steps. For the cut decision, you choose a local model or a cloud cut-brain. The cloud option may send individual still frames and transcript text to your chosen provider.
Why is cloud processing a problem for a video-editing tool?
Because your raw footage often shows more than the finished cut was meant to: unreleased customer data, internal code on screen, a glimpse into a private room in the background. A cloud service still needs that material anyway, usually for the whole processing time, and you wait for the upload to finish before anything even starts. On long recordings that is not just a privacy question, it is simply waiting time.
What does “local” actually mean in AgenticCutter?
Transcription runs with mlx-whisper directly on your Mac, the cut list is built there from the text, and rendering happens locally too. There is no step where your video has to be sent anywhere before the cutter can start working.
What stays local, and where do you get a choice?
Your video file stays on your Mac for transcription, cutting and rendering. For the cut-brain, you can choose a cloud model or a local model through LM Studio. A cloud cut-brain may receive individual still frames and transcript text. LM Studio can also run that decision locally. A YouTube upload is a separate step that you start.
How do you see what the cutter is doing?
Every cut decision shows up as a readable list against the transcript before a single frame is rendered. You read it, can ask for changes in plain language, and only after your approval does the final file get created. You drive this from the native macOS app or from Claude Code or Codex on the command line.
What does that mean for your data?
Your video and audio stay on your device as long as you do not pick a cloud model for the cut brain. There is no intermediate storage at a provider where your material sits between upload and download, and no waiting period where you do not know where your raw footage currently is.
Where does local processing cost you something?
Local processing means your Mac supplies the compute time for transcription and rendering itself, instead of offloading it to someone else's server. On very long recordings that takes correspondingly longer than a fast cloud pass, and AgenticCutter runs exclusively on macOS with Apple Silicon, not on Windows, Linux, or Intel Macs.
Frequently asked questions about local and private processing
Which data can leave my Mac?
Your video file is not uploaded for transcription, cutting or rendering. If you choose a cloud cut-brain, individual still frames and transcript text may be sent to that provider. A YouTube upload sends your finished video when you start that separate step.
Do I need internet to cut a video?
Not for transcription or rendering, those run locally. You do for the cut brain, unless you pick a local model through LM Studio.
Why is this faster than waiting for an upload?
Because the first step, transcription, starts directly on your Mac instead of waiting for a full upload to finish first. How long that takes in practice depends on your Mac and the length of the recording.
What if I want to work in the cloud anyway?
You can, by choosing a cloud model as the cut brain. AgenticCutter does not force a direction, it just makes local the default.
Related features
- Local transcription with mlx-whisper
- Choice between cloud and local cut brain (LM Studio)
- Approval before rendering
Related use case: Edit screencasts