Every image is read on your Mac.
Apple Vision does the looking — text recognition, image classification, barcode and QR decoding — and on Macs with Apple Intelligence (macOS 26+), Apple's on-device language model writes the answer from those detections. The model is text-only and never receives your image.
Screen, camera, file, or sample.
Press Option-X anywhere to drag-select a region of any display. Use a built-in, external, or Continuity Camera. Open or drop in a PNG, JPEG, HEIC, or TIFF — or start with the bundled sample image, which needs no permissions at all.
It reports what it detected.
Without Apple Intelligence, Ohms Explain shows exactly what it found — recognized text, likely subjects, decoded codes — with the measured recognition confidence. It does not invent a summary it cannot support.
A global Option-X hotkey opens a drag-select overlay on any display; the selected region is captured with ScreenCaptureKit and analyzed on-device.
Hardware & Schematics, Labels & Ingredients, Documents & Invoices, Objects & Nature, Errors & Screens — each shapes the question and the answer. Labels & Ingredients is informational only, never medical advice.
Every question and answer is kept as plain text in Application Support on your Mac — searchable, and deletable one entry or all at once.
Optionally connect Anthropic, OpenAI, Google Gemini, OpenRouter, or GitHub Models with your own key — or Ollama on your own machine. Off by default, clearly labelled, keys in the macOS Keychain.