On-device processing

What “on-device” means here

In Overshow, on-device means the work runs in the desktop application on your Mac using local CPU, GPU, and storage. Data you capture is not uploaded to Overshow for transcription, OCR, embedding generation, or search ranking. Your machine is the boundary: if a step does not need the network for the feature you are using, it does not use the network for that step.

You are always in control of what Overshow captures; processing only applies to material you have chosen to record under your current settings.

Capture and understanding

• Screen capture. Frames and derived signals are produced locally. Optical character recognition that turns visible text into searchable content runs on your Mac.
• Audio. Microphone audio you include is recorded locally. Speech-to-text uses a transcription model bundled with the app, so audio does not need to leave your machine for that step.
• Embeddings and semantic search. Vectors that power meaning-based retrieval are computed locally from your content, consistent with your settings.
• Search. Indexing and query execution run against your local database.

Ask and similar features

When you use Ask (or other assistants configured to use on-device models), responses are grounded in your local index and processed with models that run on your Mac, not sent to Overshow as a default cloud inference path for your capture content.

Offline and performance

On-device work can run while you are offline for core capture and search, subject to OS permissions and hardware. Heavy operations may slow other tasks briefly; that trade-off keeps sensitive content off shared infrastructure you do not control.

What this does not claim

On-device processing does not remove your responsibility to comply with workplace rules, consent requirements, or local recording laws. It does mean that, by default, Overshow is not sending your screen stream or raw audio to Overshow’s servers for those processing steps.