How Does Jarvis Read Text From an Image or Screenshot?

Jarvis (getjarvis.eu) reads text from images and screenshots using the vision capabilities of the models it routes to, like frontier models from Anthropic, OpenAI, and Google. Press Cmd+/ on macOS or Ctrl+/ on Windows to open the floating bar, then ask Jarvis to read or extract the text from whatever image is on screen. It handles screenshots, photos of documents, scanned PDFs, image-based menus, and text baked into graphics that you can't select or copy. Beyond plain transcription, Jarvis understands the text in context, so it can summarize a screenshotted email or pull the total from a photographed receipt. Jarvis is a screen-aware desktop assistant; free to start, then $16/month. Screenshots and images you share are processed only to answer you and are never used to train any model.

Traditional OCR converts pixels to characters and stops there. Jarvis goes further because the same vision model that reads the text also understands it. So you can ask not just "transcribe this" but "what's the deadline mentioned in this screenshot?" or "is there a phone number in this photo?" It copes with mixed layouts, columns, tables, and text overlaid on busy backgrounds better than a rigid OCR engine, because it reasons about meaning rather than matching glyph shapes. That makes it handy for receipts, business cards, slide photos, and any text trapped in an unselectable image.

Reading is usually a means to an end. Once Jarvis has the text, you can immediately ask it to do something: save a photographed address to Apple Notes, add a date from a screenshotted invite to Google Calendar, or draft a reply to an emailed image. Because Jarvis lives in a floating bar over every app and has 30+ connectors, the path from "I see text in an image" to "it's now in my system" is one short conversation, not a copy-paste-retype chore. It can also clean up messy extracted text into a tidy list or table.

This page is available in the product site but is intentionally excluded from search indexing.

Screen-aware AI