How to take a screenshot with OCR on Mac

Press ⇧ ⌘ X anywhere, or pick Take Screenshot from the Gisti menu bar icon. The screen dims, you drag a region, and the capture lands in the Stream, with its text already recognized.

Capture flow

⇧ ⌘ X – the selection overlay appears over everything, including full-screen apps. An on-screen hint bar shows the keys below for the duration of the capture.

Drag to select a region, or use one of the keys while the overlay is up:

Key	Effect
`Space`	Toggle Area vs Window mode – click a whole window instead of dragging a region (see below)
`T`	Toggle Image vs Text capture – see below
`Tab`	Copy the color under the cursor – see below
`⏎`	Capture the entire screen
`Esc`	Cancel (right-click also cancels)

While you are dragging a region, the same modifiers as in the system screenshot tool refine it. Press and hold them mid-drag – the selection reshapes immediately, without moving the mouse:

Hold	Effect
`⇧`	Lock one side – only the direction you move keeps changing
`⌥`	Resize from the center instead of the starting corner
`Space`	Move the whole selection; release to keep resizing

They combine: ⇧ ⌥ changes one dimension around a fixed center.

The two toggles are independent and combine freely: Space picks what is captured (a dragged region or a whole window), T picks what you get (an image or just its text). E.g. Window mode + Text capture OCRs one whole window without saving an image at all.
Image capture (the default): saved immediately as an element (image
- timestamp) so it shows up right away, then OCR runs in the background and attaches its text a moment later.
Text capture: OCR runs immediately, the recognized text is copied to the clipboard and saved as a text element. No image file is written to disk in this mode. If no text is recognized, Gisti beeps and the clipboard is left untouched.

Loupe and color picker

A magnifier follows the cursor while you select, showing the pixels around it with the exact one you are on outlined in the center. The patch of screen it is reading is outlined at the cursor too, so what is in frame is visible without looking away from the pointer. Use it to line a selection edge up against a pixel, or to read a color. Scroll to change the magnification – it steps between 4×, 8× and 16× and remembers the last one. Turn it off in Settings → Screenshots → Show loupe if you would rather see the screen unobstructed; the size and coordinates stay, and so does the color picker below.

Tab copies the color under the cursor. It also lands in the Stream as a text element, so you can find it later by searching for its own digits – that’s the part Digital Color Meter doesn’t do. The overlay stays open, so you can pick several colors in a row before capturing (or press Esc if a color was all you wanted).

The notation follows the format picked in the screenshot viewer – HEX, RGB or HSL. Both eyedroppers share one setting, and it survives a relaunch.

The size and coordinates under the magnifier are in pixels by default, so the number you drag to is the number of pixels you get: on a Retina display a 400×300 pt region is an 800×600 px file. Switch the unit to pt in the viewer’s status bar and this readout follows.

The magnifier shows the screen as it is underneath Gisti’s own overlay, so the dimmed selection area and the hint bar never tint what you are reading. It refreshes as the pointer moves, and about once every couple of seconds while you hold still – so a video or an animation under a stationary cursor can lag by a moment.

Window mode

Space switches the overlay to Window mode – the window under the cursor is highlighted, one click captures it. The window is captured as an object, not cut out of the screen:

Rounded corners stay transparent in the saved PNG – no desktop pixels baked in, so the shot pastes cleanly onto any background.
The window is captured whole, even when it’s partially covered by other windows.
The menu bar (or a single status-bar icon) can be clicked and captured the same way.

If there’s nothing capturable under the cursor, the click just beeps.

OCR

Runs locally via Apple’s Vision framework – no cloud, no upload.
Recognized text is indexed for search – find a screenshot by a word that was on the screen (and semantically, see Semantic search).
Copy OCR Text (⌘ R) extracts it as plain text.
OCR always runs on capture, for both Image and Text mode – there’s no setting to turn it off.

After capture

Open the element to view, crop, rotate, or pick colors.
⌘ Y summarizes what the screenshot says.
Settings → Screenshots → Open after capture opens the panel and jumps straight to the screenshot viewer right after you take a screenshot, instead of just leaving it in the stream (on by default).
Settings → Screenshots also sets the hotkey and the image storage limit (shared by screenshots and images copied to the clipboard).

Copying at 1×

A Retina capture holds four times the pixels of what was on your screen – a 900 × 600 pt window is 1800 × 1200 px in the file. That detail is worth keeping on disk, but it isn’t free when you paste the shot into an AI chat: image models price pictures by area, so the Retina copy can cost up to four times the tokens of the same screenshot at 1×, for content the model reads just as well.

Settings → Screenshots → Copy Retina screenshots at 1× (off by default) hands the smaller version to the clipboard and leaves the stored file at full resolution. With it on:

⇧ ⌘ X puts the 1× version on the clipboard the moment a capture finishes – no extra step before you paste.
⌘ C copies at 1× on a screenshot selected in the Stream and in the viewer, crop selections included.
Dragging a screenshot out of the Stream drops the 1× bitmap.
Every one of those says Copied at 1× instead of Copied image, so you always know which version you are holding.

What it deliberately leaves alone:

The stored file. Export, Share and Open in Preview give you the full-resolution screenshot – and so does dragging into Finder or any other target that takes the file rather than the pixels. Shrinking is for hand-offs that are context for a model, not for deliverables.
Anything Gisti didn’t capture. Images copied from the clipboard, sent in through the Share extension, or clipped by the browser extension are copied untouched. Gisti records the display scale at the moment it takes a screenshot and won’t guess it from pixel dimensions for a picture that arrived some other way.
Screenshots from earlier versions of Gisti, which don’t carry that recorded scale, and screenshots from a non-Retina display, which are already at 1×.
Screenshots you resized by hand in the viewer. Cropping and rotating keep the recorded scale; resizing retires it, because once you have picked a size yourself the picture is no longer a 2× rendering of a screen region and halving it again would be second-guessing you. Reset brings the scale back along with the original picture, and so does undoing a Reset.

Full-screen captures save less than region captures. Past a certain size every provider downscales an image on its own before the model sees it, so a 2× full-screen shot was already being shrunk – the saving is largest on the window and region captures you take most often.

Overriding it for one copy

⌥ ⌘ C copies the other way round from whatever the setting says: full resolution when the preference is on, 1× when it’s off. It works on a screenshot selected in the Stream and in the viewer. It flips the preference and nothing more – it still won’t shrink a picture Gisti didn’t capture.