Images

An Image is a picture you create from a prompt or bring in yourself, then refine until it's right. Images live under the Media filter group in your Artifacts, so you can find every picture you've made in one place.

Create an Image

You have two ways to start:

• Generate from chat. Describe the picture you want in the chat and send it. MotherFluxer creates an Image artifact you can open and keep refining. Where the model options are shown, choose the one that fits your needs — each lists its cost and speed. The picture shows up in the chat and in Artifacts the moment the model finishes, already named and ready to open; its small preview tile follows a few seconds later.
• Upload your own. Bring in a picture you already have and treat it like any generated Image: open it in the editor and refine it with the same tools.

Every Image you make is saved to your Artifacts and can be organized into Projects.

Open the editor

Open any Image to enter the editor. Its tools sit across the top of the screen, ready to use — there is no mode to switch into first:

• The tool menus on the left — Add, Canvas, Layers and File. The top bar only puts things in and holds what belongs to the whole picture: Add brings in something new, Canvas is the picture as a whole — canvas size, background, the zoom range, the grid, and the picture's own rotate, straighten, crop and flip — and Layers is the stack of everything already there.
• The bar along the bottom — the settings of whatever you have selected. Click a thing on the canvas or in the Layers list and its controls appear there; there is no menu to choose first, because picking the thing is the choice. Click the tab you are already on to deselect, and the bar goes away with it.
• Smart Edit — the MotherFluxer logo, first on the left. Describe a change in words and let MotherFluxer make it.
• File — the last menu on the left, after Layers. Save, restore, share and export versions of the Image.

Adding something always works the same way: the menu closes behind the add, the new thing arrives selected, and its settings are already open along the bottom — so you dial it in right away, with the menu out of the way.

The editor draws your picture on your device's graphics processor, as every editor of pictures does; where a device can't run it, opening an Image says so in place of the editor. The Image's card still shows everywhere. If one of the pictures in the Image can't be loaded, the editor says so at the centre of the screen — Image failed to load, with Reload — rather than leaving a blank where the picture should be.

Smart Edit an Image

In Smart Edit, you describe the change in Discuss and MotherFluxer works out which edit to run. Images have the widest set:

• Change one part of the picture — add something to a region, remove an object, replace or recolour what's there, or expand the edges outward.
• Change the whole picture — restyle it into another medium, restage the same subject in a new scene, or match the style of a reference image.
• Change a person — swap a face, try on a hairstyle or outfit, or adjust an expression.
• Make it better as it is — relight, upscale, restore, colorize, or auto-enhance.
• Tune what's already there — Change settings sets an item's look, place, edge or copies from a sentence, without repainting the picture.

A few notes:

• Some edits need a painted region. When the AI suggests one, the card reads Set up & generate — paint over the area you want changed, then generate, so MotherFluxer knows exactly where to work.
• The area lands where you marked it, on any page of the Image and wherever the picture sits — moved, resized, turned or cropped. Two cases are refused before anything runs: an area that misses the picture altogether (The area you marked is not on the picture. Mark an area on it and try again.), and a picture tilted with perspective (Area edits do not work on a picture with perspective. Reset its perspective and try again.).
• Large photos work on a phone too. An area edit on a 48-megapixel photo runs on the phone that took it, and everything outside the area you marked stays exactly as it was. If the area you marked is itself larger than the edit's AI model takes, your device scales it down to fit before it is sent and says so for a moment — The area you marked is larger than this model takes, so it was scaled down to fit. — so that area can come back a little softer than the rest of the picture.
• A flatten keeps everything you placed on the picture, wherever you start it — from the Smart Edit panel, from an edit the AI suggests in a chat, or as part of a chain of edits. The page is baked into one picture first, so the text, shapes and drawings over it end up in the result. In the editor that is the page you have open; from a chat, where the Image shows as its first page, it is the first page. In a chain, the flatten is always the first step, and every later step edits what it made.
• The expression controls take effect. When you adjust an expression, each slider — the gaze left and right or up and down, the eyes, the eyebrows, and the head's tilt, turn and roll — changes the result by the amount you set.
• Edits that work on a face are not offered in Illinois or Texas. Swapping a face, trying on a hairstyle or outfit, and adjusting an expression all measure the face in the photo, and those states require the person's written permission first. There, the three edits don't appear among the edits you can run, and one the AI suggests won't run.
• If the edit you want doesn't exist for Images yet, the AI says so rather than running something else.

The edits Smart Edit can run on an Image, each with what it does:

• Remove object — Erase something and fill the gap naturally
• Add to region — Paint a region and describe what to put there
• Replace in region — Swap the selected thing for something else
• Recolor region — Change the color of the selected thing
• Expand edges — Extend the image beyond its current edges
• Background — Decide what sits behind the subject — cut the background out to transparency, fill it with one flat colour, or generate a whole new scene behind it
• Describe edit — Tell the AI what to change in plain language
• Match style — Match the look of a reference image
• Add text — Render new text baked into the scene (a sign, a label, a storefront)
• Edit text — Change text that's already in the image (keep the scene around it)
• Swap face — Replace the subject's face with a face from another photo
• Try on — Try on a new hairstyle or a new outfit — pick one and the subject keeps their identity
• Adjust expression — Re-pose a portrait's gaze, blink, eyebrows, and head angle
• Combine refs — Combine subjects from one or more reference images
• Relight — Change the lighting direction, mood, or temperature
• Upscale — Increase resolution and sharpen detail
• Restore — Repair a degraded, noisy, or blurry image
• Colorize — Add color to a black-and-white image
• Auto-enhance — Make colors, contrast, and lighting pop — keeps the content and medium
• Realify — Turn a drawing, painting, or render into a realistic photo — keeps the composition
• Change settings — Change an item's settings from a sentence

The Image hands-on tools

You do not add a look from a menu at the top — you pick the picture you want to change, and its effects are already there along the bottom. Select a picture and it has its own menus: Image, Effects, Border and, last, Transform — text, a shape or a drawing has its own settings menu in place of Image, named for what it is. Transform is the same on every item, four rows at most: a first row of Crop, flip and, on a picture that can be cut, Splice (text, a table, a mask and a blur have no Crop; Splice opens the ways to cut in the menu's place, with a row at the top to come back), then Perspective — press it to bring the tilt controls up on the picture (the row reads On screen while they are up) and again to put them away; otherwise it shows how far the item leans, with a reset while it is tilted — then Opacity, each on a row of its own, and at the bottom where it sits, its size and its turn, written as plain numbers with their unit (412px 236px), with the size's lock between its two numbers. An item with no tilt, such as a mask or a 3D model, simply has one row fewer; a video clip, whose own settings take a row, has Perspective and Opacity side by side. Effects and Border work as every editor's item menus do: a strip of cards in sections, the picture's layers in a container at its right end, and each layer's Recipe. A Weather layer stars the same way, its sprite included, and its Starred card puts that weather back on a picture or a group; a drawn mark holds no weather, so its Starred section leaves those cards out. A mask is an item of its own, so it comes from the Add menu at the top like everything else you put down, and its first menu is Mask rather than a settings menu. Splice cuts a picture that is already there, so it lives on that picture's Transform tab.

The Effects strip is one list in nine sections — Adjust, Beauty, Filters, Styles, Distort, Light, Weather, Depth and Form, followed by a Starred section for each kind you have starred — sorted by what the effect does to the picture rather than by how it is built. Picking a card in any section puts that effect on the layer you are working on, and Recipe opens its settings in place of the cards; every tile previews on your own picture. What each section holds, and every setting an effect's Recipe has, are on the visual effect page.

A few cards open a setting of the layer rather than adding a look. The last two in Adjust are Mark, a circle, a box, an underline or an arrow drawn round the layer to point it out, and Blend, how the layer lands on what is beneath it; Form holds Repeat and Deform. Each card wears a dot while the layer carries that setting and opens its Recipe; the mark, the blend and the copies Repeat makes each have a page of their own in these help docs. The Starred sections keep each kind apart — your effects, a tuned weather among them, then your blends, marks and copies — each named for its kind, and the search at the strip's start looks through all of them, and through what the community has posted.

An edge is not an effect of the picture: it belongs to one layer, and the Border menu is where it is drawn.

The Add menu is where everything new comes from, and it is a grid of equal tiles — Image, Freeform, Shape, Text, Table, Emoji, 3D Model and Mask, with Browse standing tall at their right. Every tile shows a small picture of what it adds rather than its name — the little landscape is Image, the red squiggle is Freeform, the serif "Abc" is Text, the gold cube is 3D Model, the glasses-and-mustache disguise is Mask, the globe is Browse — and a screen reader still reads out each tile's name. Text, Shape, Emoji and Table put their element down immediately — selected, with its settings open along the bottom; Freeform starts a drawing session; Image adds your newest picture and 3D Model your newest 3D model, as an object you can restyle and pose in place — each at once, and each swapped for another of yours from its first tab along the bottom; Mask puts a mask down in the middle of the picture; Browse opens the places to look for a picture — an upload from your device (a picture, a 3D model or an SVG file), your artifacts, and Starred and community, a chooser of the items and groups you have starred and the community has posted, by kind, where a pick places the whole thing with its settings on it. Pictures other people have published are in the search on an Image strip, under Community, a starred effect is in the search on Effects, under Starred, and a cut shape or an edge you starred, or one the community posted, is in the search on Border. The tiles in turn:

• Text — words on the picture, typed in place, with a shape behind them if you want one and a look picked from a strip.
• Shape — an outline from a library of eighteen, reshaped by its rings, its fill and line its ink.
• Freeform — a drawing of strokes, painted freehand or clicked out as a connector: pressing it puts an empty drawing down, selected, and you are drawing straight away. Each brush's Recipe sets how it lays its Stamps, how its line tapers along its length or with a pen's pressure or speed, and how it Lands — any brush can be set to Erase.
• Emoji — an emoji, drawn in 3D or flat.
• Table — a styled grid of rows and columns you type straight into. It arrives as an empty 3×3 table, selected, with its rows, columns, borders, header and stripe controls open along the bottom.
• Image — adds your newest image at once, selected, with its Image tab open along the bottom. That tab shows the reel of your own images, opening with a search column that narrows it as you type, and at the right end the picture it holds now: pick another and it takes the place of the one on the picture, keeping where it sits. If you have no images yet, Browse opens on its upload instead.
• 3D Model — adds your newest 3D model at once: it lands on the picture as an object, selected, with its 3D tab open along the bottom, where you can switch it to another of your models, pose it and change its surface.
• Mask — press the tile and a mask lands in the middle of the picture, selected, wearing the library's default look and following nothing; having it sit on a face is a choice you make from the target button at the bottom of its card.
• Browse — an SVG file you upload here comes in as things you can edit rather than as a flat picture: shapes, drawings, pictures and text, grouped where the file groups them, selected, in one step Undo takes back. Whatever the file holds that these can't, the editor names once, Brought in. Left behind: and the list; a file it can't read says That file isn't one this editor can read., and one too big to bring in says That file is too large to bring in.

The Layers menu is the stack itself — every drawing and picture in one row of cards, front to back:

• Every card is a picture of your own image, not an icon — your picture with only that layer switched on, and zoomed in on that layer so a small mark still reads at card size. The looks and corrections a layer carries show in its card, because they are settings on that layer rather than layers of their own. A group shows the layers it holds, composited, with a small stack of cards peeking out behind it — one for each extra layer, up to three.
• A card whose layer draws nothing shows its name instead, and a layer your crop has cut away shows an empty card — which is exactly what the crop did to it.
• Selecting a card focuses that layer: its box appears on the picture and its settings open along the bottom, while the list stays open, so you can walk the stack card by card. The list closes when you press its tab again, open another menu, or click anywhere outside it.
• Selecting a card also brings up its controls over the picture — its buttons along the top, and arrows along the bottom to move the layer forward or back, which is what decides what sits in front of what. The picture darkens while its card is selected, so those controls stay readable even over a white layer. An unselected card is just the picture.
• A locked layer's card shows a small lock at its left edge, selected or not, and has no delete button. Try to move a locked layer on the picture and a lock appears over it for a moment, then fades.
• A group moves through the stack as one piece. A layer or a whole group stepping toward a group jumps past the whole of it in one move, never into the middle of its layers, and a layer inside a group moves only among that group's own layers. An arrow that would change nothing is dimmed.
• The list also lets you put several layers into a group that moves, turns, flips and resizes as one.
• The bottom of the list is the picture everything else sits on. It's an ordinary layer like any other — select it, move it, reorder it, group it, or delete it. Lock it if you'd rather it stayed put.

The Canvas menu is the picture as a whole, in four rows — first the picture's own edits, rotate, straighten, crop and flip, then the canvas size and background, the zoom range, and the grid. Each of the last three rows is split into sections you press to open in place: Canvas shows the size beside Background, Zoom shows the range beside Default, and Grid works the same way. How many pages the Image has is shown by the page controls on the screen. Once you resize the canvas, a Reset appears right after the size when you open Canvas; press it to put the canvas back to the picture's own size as it is now. Crop and Straighten put their handles on the picture, and their buttons stand in a small container just under it (or above it, when the menu along the bottom would cover it): a cross to cancel and a highlighted check mark for Done — Straighten adds a reset, and its angle dial stays at the bottom of the screen. An item's own crop, from Crop on its Transform menu, finishes the same way. Background is what shows wherever the picture is transparent: leave it transparent and the editor shows a dark grey checkerboard there — the same one the Video editor uses, whichever light or dark look the app is in — or pick a colour to fill those areas: open Background and press its swatch. The swatch's Transparent background control clears it again, and dragging through colours in the picker is one step of Undo. A width or height you type applies when you press Enter or leave the field, not while you are still typing, and the Up and Down arrow keys step it one pixel at a time, or ten with Shift. Each resize is one step of Undo, and so is a run of arrow presses.

The picture's own settings

The Canvas menu's rows set the picture as a whole: its own turn, straighten, crop and flips, its size and its background.

Canvas

• Crop — The part of the whole page the picture keeps, as shares of its width and height.
• Rotation — How many quarter turns clockwise the whole page stands at, before it is mirrored.
  • Upright · Quarter turn · Half turn · Three-quarter turn
• Straighten — How far the whole page is turned to straighten it, in degrees either way, beside its quarter turns.
• Flip horizontal — Whether the whole page is mirrored left to right, after it is turned.
• Flip vertical — Whether the whole page is mirrored top to bottom, after it is turned.
• Canvas — The page's size in its own pixels, width then height: every item on it is scaled with it.
• Background — What shows wherever nothing on the page is drawn: a colour, or nothing at all.

The buttons over a selected item

Select anything in the Image — a picture, text, a shape, a drawing, a table, an emoji, a mask or a group — and a short row of buttons stands just above it (just inside its top edge when it fills the screen), the same row on every item:

• Turn, first — drag it to turn the item; a press that doesn't move does nothing. While you drag, the other buttons and the grips step aside and the status at the bottom centre reads the angle, such as Rotate · 37°. The turn settles gently onto 0°, 90°, 180° and 270° as it passes, and holding Shift steps it 15° at a time. From the keyboard, the left and right arrow keys on it turn the item a step at a time.
• Duplicate — makes a copy of the item.
• Delete — removes it. It sits a little apart from the others, so a reach for Duplicate never lands on it.
• More — everything reached for less often, always in the same order: Lock (or Unlock), then Fit display, then a row of four arrows for the item's place in the stack — bring to front, forward one layer, back one layer and send to back, front on the left, the same marks the Layers menu uses. An arrow that would change nothing is dimmed.

Fit display makes the item, at its own proportions, as big as fits whole inside the picture — the part your crop keeps — centred on it, so whichever side reaches the picture's edge first sets its size. Only its size and place change: a turned or perspective-tilted item stays turned and tilted, and it's the box you see around it that fits, while its flip, crop, opacity and effects stay too. A text's words grow or shrink with its box. A table grows only in width, since its height follows its rows; upright, it fills the picture's width. It's a one-time move, and one Undo puts the item back.

A locked item keeps only Duplicate and More, with Unlock in More; it can't be turned, moved, resized or deleted until you unlock it, and Fit display isn't offered. Locking a group locks every layer in it, and a group reads as locked while any of its layers is. A mask that follows something is placed by what it follows, so it isn't offered Fit display either.

Find what's under a spot in the picture

A tap or click on the picture picks the topmost thing you can actually see at that spot, and the same press can drag it straight away: once the pointer travels a few pixels the thing moves, while a slightly shaky tap only selects it. Only the ink counts: clicking through the see-through corner of a cut-out, or through the middle of an unfilled shape, reaches whatever is behind it.

Tap the selected thing again and the selection steps to whatever is underneath it at that spot — so anything hidden behind something else is reachable without leaving the canvas — or types into a selected text box or table, as on every editor. Dragging the selected thing moves it instead of stepping down. A layer that belongs to a group selects its whole group first; tap again and you step into the group, to the layer under your finger.

The picture you started with is a layer like any other: click it to select it and open its menus along the bottom. Because it sits under everything else, clicking a spot where nothing else is drawn selects it. Clicking outside the picture, where nothing is drawn at all, deselects.

To see everything stacked at one spot at once, press and hold on it. A crosshair marks the spot, the Layers list opens showing only the layers at that point, and a blue Filtered card sits at the front of the list. Press its ✕ to go back to the full list; closing the Layers menu clears it too.

Selecting a layer from the list opens its menus along the bottom, so you land on its controls rather than having to go find them — and the list stays open while you do.

Cut an object out with Splice

Splice lifts a single object out of a picture and turns it into its own layer you can move, scale, and rotate. It works on any image in your artifact — the background picture, or any image layer you've placed on top of it. It is the Splice cell at the end of the selected picture's Transform tab's first row, along the bottom: pick the picture, open Transform, press Splice, and the ways to cut stand in the menu's place under a row that takes you back. Object is the tap-to-select cut described here; Depth slices the picture into pieces by how far away each part of it is — and it shows you what it measured before it cuts (see below). The menu stays open for the whole cut; once the pieces exist they behave like any other layers, and the new piece's Transform tab is what you see.

1. Press Object under the Splice row. It stays lit while you cut — press it again to stop — and the image you picked comes to the front while everything else fades back, so you can see exactly what you're working on.
2. Wait a moment while it's prepared, then tap the thing you want to cut out, or drag a box around it. MotherFluxer traces its outline and dims what you're leaving behind.
3. If the outline caught too little or too much, refine it with the buttons in the small container under the picture. Add (+) means your next tap brings more into the selection; Subtract (−) means your next tap takes something out, and the hint beside them says which is on — Tap to add or Tap to remove. Tap as many times as you need — each tap re-reads the whole selection, so it keeps getting closer. Undo steps the selection back one tap at a time, then the box, without re-reading anything.
4. Press the highlighted Done (✓) to make the cut. Cancel (✕) closes the tool and changes nothing, leaving the image selected.

Splice runs on your own device — a computer, a tablet or a phone — so the first cut after opening the editor takes a moment longer while it gets ready. If your device can't run it, the editor says so plainly: Splice couldn't load on this device.

A very large photo cuts on a phone too — 48 megapixels and more, HDR photos included: each piece is made and saved a strip at a time at full resolution, and an HDR photo's pieces keep its bright highlights.

Cut a picture into near and far with Depth

Depth cuts the whole picture at once, into pieces sorted by how far away each part of it is. Pick Depth as the way to cut and it reads the picture — that takes a few seconds — and then, instead of cutting straight away, it shows you what it found:

1. A preview of the picture with the reading drawn over it. Brighter is nearer, and each shade is one piece, so you can see exactly where the cuts will fall before anything happens.
2. A Pieces slider — anywhere from two to six. Drag it and the preview re-shades immediately; nothing is re-read and nothing is cut yet.
3. Cut makes the pieces. Leaving the step instead throws the reading away and changes nothing.

Each piece arrives as its own layer, in front-to-back order, already carrying how far away it is — which is what the group's distance settings use to add fog, focus and lighting across the set.

A Depth cut of a 48-megapixel photo into six pieces works on a phone as well. A page holding that many very large pictures keeps each one at the detail the screen shows, so when you zoom far in, the part you are looking at sharpens a moment after you stop.

What happens to your picture after a Splice

Every cut works the same way, whichever picture you cut — the background picture included. The picture is replaced by two genuinely separate layers: the object you cut, and the rest of the picture with a transparent hole where the object used to be. Each piece has a soft edge so it doesn't look pasted on, both sit exactly where the original was (so nothing appears to move until you move it), and the pair arrives already grouped.

Because the two pieces are real, independent images:

• Moving either piece never brings back what was cut — the hole travels with the remainder, and the object carries only its own pixels. The pieces stay separate forever.
• Undo puts the original, un-cut picture back in a single step.
• Duplicating a piece copies just that piece — you get a second copy to place.
• Deleting a piece removes that piece, exactly like deleting any other layer. To restore the full picture, use undo instead.
• What you tracked goes with it. A face or an object you had tracked on the picture moves to the piece it stands on — the cut-out when you cut it out, otherwise the piece that holds it — so a mask following it keeps following it, and a new mask can attach to it there. Nothing is read again.

The menu offers exactly the pictures a cut can work on. Pictures that are locked or have had perspective applied aren't offered, and neither is a 3D object — a cut replaces the picture with its pieces, so a locked picture stays untouched until you unlock it, and a 3D object is redrawn from its model every time you leave its 3D and Transform tabs with a change.

Set how far you can zoom

Open Canvas and the Zoom row sets how far the view is allowed to go. 100% means the whole picture fitted to the window. The number before to is how far out you may go: zooming out under 100% shrinks the page in place, centred, down to this number. The number after to is how far in you may go, and Default on the right is the view at rest: where the picture opens, where the reset button in the top bar puts you back, and where a wheel or pinch pauses as it passes — keep scrolling the same way and it carries on through; scroll back and it moves at once.

Out of the box an Image goes from 25% up to 500%. You can push the ceiling up to 2000% — worth doing on a very large picture, where reaching one screen pixel per picture pixel takes far more than 500% — and set the lowest point anywhere from 10% to 100%.

You do not have to widen the range first: type a Default outside it and the range opens up to include it.

The zoom range is remembered for each Image on its own, the same way the grid is. Every Image starts on 25%-to-500% with fitted as its default; once you change the range on an Image, it is there exactly as you left it the next time you open that Image — through every new version too — while Images you never changed it on keep the default. Opening an Image always shows the whole picture fitted; the Default is where the reset button takes you. The range is yours rather than part of the picture: someone you share the Image with has their own, and it never appears in anything you export.

Zoom and pan with two fingers

On a phone or a tablet, put two fingers on the picture and spread them to zoom in or bring them together to zoom out; the picture grows and shrinks around the point between your fingers, the same way the mouse wheel zooms around the pointer. Keep both fingers down and drag to pan. Two fingers only ever zoom and pan — they never start a tool, draw a stroke, pick a layer or open the picture's perspective handles, and a pinch that begins on a layer leaves it unpicked — and one finger goes on doing exactly what it did: a tap picks, a drag moves what you picked or paints with the tool you have open. Tap the picture with two fingers to undo your last change, and with three fingers to redo it, the way a drawing app does. Lift one finger and the gesture ends where you left it; the picture stays zoomed until you zoom again or press the reset button in the top bar. The page itself never zooms, so the menus and bars stay put while the picture moves.

Very large pictures zoom the same way. A photo straight from a modern phone camera — 48 megapixels and more, HDR photos included — opens, zooms all the way in to the editor's ceiling and pans at full detail on the phone that took it, because the editor only ever draws the part of the picture that is on screen rather than the whole picture at once. The same is true on a laptop: zooming into a large picture costs the same whatever its size.

A few effects rearrange the whole picture rather than each part of it — Kaleido, Spiral, Pixel Sort and a strong Fisheye — so they still draw the whole picture at every zoom. On a phone, a large picture wearing one of them looks a little softer while you are zoomed far in, which keeps the editor within the phone's memory; your export always has full detail.

Line things up with the grid

Open Canvas and find the Grid row. It has four parts side by side: Grid, Snap, Format and Color. Press Grid to turn the guide lines on — it then reads On — and press it again to turn them off. While the grid is off, the other three stay in the row but are greyed out until you turn it on.

Pressing Snap, Format or Color opens that part in its own place with all its settings and an ✕ at its end; the other parts shrink to just their names, so you can jump straight to one of them, and the ✕ puts the row back.

Snap says what the grid does to the things you move — Off, Soft or Hard. Open it and it shows two settings, each reading its current value: press the first to turn the snap on or off, and the second to switch between:

• Soft — anything you move nearby is gently pulled onto a line.
• Hard — things can only come to rest exactly on a line.

A new grid is a visual reference only — its snap starts Off — and the grid only pulls while it is on: lines you can't see never pull anything. Your choice of Soft or Hard is kept while the snap is off (it shows greyed out), so turning it back on gives you the same pull as before.

Format sets how the lines are made: Gap is the space between lines and Weight how thick they are. Color sets their colours: Lines for the grid itself and Center for the cross through the middle. The grid is centred on the picture — a line runs straight through the middle in each direction — and it stays centred on its own: resize the picture or change the gap and it still is, which is the part you would otherwise have to work out by hand. The lines are drawn on whole screen pixels, so they stay the same crisp width at any zoom.

Center draws a cross through the middle of the picture. It is off until you pick a colour for it — pick one and the cross appears in that colour, a little heavier than the grid lines so it reads as the middle; the ✕ beside it turns it off again. The cross uses the grid's own snap while the grid is on — turn the grid off and the cross stays visible but stops pulling anything.

Once a grid is on, it applies as you work rather than only afterwards: a shape, text box, emoji or picture lands on the grid the moment you add it, and a selection you draw for Smart Edit, a connector's points and the crop rectangle all line up to it while you drag them. Freehand brush strokes are left alone, since a brush stroke is wherever you painted it.

Two things help while you work:

• Hold Command (or Control on Windows) while dragging to ignore the grid for that one movement, without turning it off.
• Arrow keys nudge the selected thing one pixel, or ten with Shift held. With the snap set to Hard, an arrow key moves it a whole square at a time and puts it on the line — so you can line things up from the keyboard alone.

The grid is remembered for each Image on its own. Every Image starts with the grid off; once you change it on an Image, it is there exactly as you left it the next time you open that Image — through every new version too — while Images you never changed it on keep the default. It is yours rather than part of the picture: someone you share the Image with has their own grid, and it never appears in anything you export.

Work with multiple pages

An Image can hold up to 10 pages. The page controls sit in the bottom-right corner of the editor at every zoom, whenever nothing is selected — they fade out while you work on an item and come back when you deselect:

• The ‹ and › arrows, either side of the page number, such as 2/3 — the arrows take you to the page before and the page after.
• Delete page removes the page you are on. It appears once the Image has more than one page.
• Add page (the +) opens two choices: Blank page — a new page right after this one, the same size and background as the page you are on — and Copy of this page — a copy of the page you are on, placed right after it.

Zooming never moves the page or its controls. Each page is edited independently, with its own Undo. If a page can't be read when you turn to it, the editor says Image failed to load and keeps you on the page you were on, with that page's saved work untouched — turning to it again tries once more.

Undo and Redo

Every change is reversible. Use Undo to step back and Redo to step forward again, so you can experiment freely and never lose a version you liked.

A number you type into a layer's Position, Size or Rotation applies when you press Enter or leave the field, and each value you type is its own step of Undo. The Up and Down arrow keys step a field, ten times as far with Shift, and a run of arrow presses on one field is a single step. Open Position and a small square beside its numbers picks which corner of the item they read — press it to step round the four corners. A Size you type keeps that corner where it is on screen, even on a turned item, and the box grows or shrinks away from it; a text's words scale with its width.

Your picture is saved as you go, a moment after you pause, so you don't have to close the editor to keep a change.

If the same Image was changed somewhere else while you had it open — another tab, another device, or someone you share it with — the editor stops saving instead of writing over that work, and tells you the Image was edited elsewhere, with Reload to pull in the newer version.

Export your Image

When you're happy with the result, open the File menu to export a full-resolution PNG. That gives you the highest-quality version of your Image, ready to download and use anywhere, in a file named after the Image. An Image holding an HDR photo exports as an Ultra HDR JPEG instead: it keeps the photo's bright highlights on an HDR screen and looks right on any other. If part of that Image is see-through, it exports as a PNG, because a JPEG cannot hold see-through areas.

Export lists Download beside Add to an App…, which puts the same file into one of your Apps' Assets instead of downloading it.

When AI made any part of the Image, the exported file carries a standard, machine-readable tag inside it saying so — nothing you can see in the picture, but readable by the apps and sites that check for AI-made content. A photo of your own that no AI step has touched carries no tag.

A very large photo exports at full resolution on a phone too — the file is written a strip at a time rather than all at once, so a 48-megapixel export finishes on the phone that took the photo.

The export is drawn by the same graphics engine that draws the preview, so what you see on screen is what you get in the file — and an Image carrying heavy effects like Comic or Watercolor exports in seconds rather than minutes.