Making images

The Assistant’s prompt floats at the bottom of a Canvas (press / to start typing in it). Hover it and a small handle appears: above it while it floats, just to its left while it’s docked. Drag the handle to the right edge, or press it and choose Dock right, to keep the whole conversation beside the Canvas; Float puts it back. It remembers where you left it. To add images from your computer, use Add images or videos (+) in the toolbar. Choose Image from the mode button at the start of the prompt (see Talking with the Assistant; in Ask the Assistant talks it through with you first), describe an image and press Enter: the prompt clears, ready for your next one, and the image’s place appears in the middle of your view straight away, glowing softly, like light behind glass, while it is generated, and the picture fades in when it is ready, usually within a minute. Meanwhile it is an ordinary image: move, resize or delete it, and the result lands where you put it. If you reload while it is being made, it carries on and still fills its place. If the Assistant can’t take a prompt at all, the prompt comes back with why (or, if you’ve already started the next one, why is shown without touching it). If it can’t be made, it stays in place, its glow turned red and still: hover it for Retry, or select it to see why in the selection toolbar at the top right, with Retry (delete it as any node). To work from an image, right-click it and choose Use as reference (or More in its toolbar); select several first to choose them all at once. Or drag it onto the prompt: the prompt lights up when it can take what you’re dragging, and the image stays where it was on the Canvas. You can also paste an image (a screenshot, say) into the prompt with Ctrl/⌘ + V: it becomes a reference without being added to your assets or the Canvas, and pasting the same image again adds nothing. If the model you’ve chosen can’t take that many images, the prompt says so and nothing is added. Selecting an image on its own never adds it to the prompt. With one image, the Assistant changes it: the changed image is a new image beside the original, which is left as it was. If there’s no room beside it in your view (the original is at the edge, or off screen), the new image lands in the middle of your view instead: what you make always arrives where you can see it. With several, it makes a new image from them; refer to them in your prompt by the names shown above it (Image 1, Image 2). Each one is shown above the prompt as it will be sent: remove any you don’t want with ×, which leaves the Canvas as it is. Sending takes them with the prompt; choose them again for the next one.

To try an image another way, right-click it and choose Iterate: a variation starts straight away, with no prompt, and appears beside the original, which is never changed. A variation keeps the subject, style and mood, and shows it from another angle, arrangement and light, like another frame from the same shoot. It always uses Iterate’s Auto model (FLUX.3 Klein), and leaves whatever you’re writing in the prompt alone. To say where to take it instead, choose Iterate with prompt…: the prompt switches to Image and the image goes into it as the one to work from (the prompt says “Iterating on Image 1”), ready to type in. It stays there while you select other things, until you remove it with × or send. The result is a new image beside the original, and you can choose another model as usual.

Settings

The sliders button beside the price opens Settings: the Model, then the settings that model offers. For an image, its shape (Aspect ratio) and Size, and for a model that has them, its Quality; for a video, also its Resolution, Duration and Sound. Hover the button (or listen to it with a screen reader) to hear what is set, such as Seedance 2.0 Fast · 16:9 · 5s. The price beside the send button follows what you set. Settings are remembered in this browser: the model you chose for Image and for Video, and what you set for each model. A model you haven’t used before starts from its own recommended settings. Auto shape takes the shape of what you make it from (an image you edit or animate, or the first image you chose), as near as the model makes it, and never crops it; a shape you choose stays, whatever images you add or take away. In an image’s editor a change keeps the image’s shape and resolution, so only the model can be chosen there, and its Quality where it has one.

Quality is offered only by models that have quality tiers, such as GPT Image 2.5 (Low, Medium, High, XHigh and Max, starting at High). A higher tier takes longer and costs more, sometimes many times more: the price beside the send button follows the tier you choose, so check it before you send. Each image you send with it (the one you change, and each reference) adds a little too.

The Assistant picks a model for what you ask: Settings show Auto, and the button names the model Auto is using. You can choose another in Settings, which lists the models that can do what you’re asking, or name one in the prompt itself with --model, for example a red chair on grass --model flux-klein. The --model part is taken out before the prompt is sent, and Settings show the model you named; a name that isn’t a model’s, or a model that can’t do what you ask, is refused with why. Each model’s page lists its name (see Models). FLUX.1 Schnell (--model explore) makes Explore’s quick drafts that your plan includes each day (see Explore). Its price is shown beside the send button before you send (see Usage and prices). What the Assistant makes is an image like any other: select, move, resize, copy or delete it, and undo placing it. Every image it makes or changes is kept in your Space, with the prompt and what it was made from.

Editing an image

Double-click an image on a Canvas, or right-click it (or use More in its toolbar) and choose Edit, to open it in the image editor. The image grows from where it was on your Canvas into the middle of the screen, the Canvas’s tools give way to the editor’s, and the address becomes the image’s own (/images/…). Your Canvas stays just behind it.

The Assistant stays too, docked beside the editor with the image’s own conversation (see An image’s own conversation), and while the editor is open it is about the image in front of you. In Image mode the prompt reads “Describe what you want to change…”; in Video it animates the image instead, beside it on the Canvas; in Ask you can talk the change through first. Say what to change (“make it night, with the lights on”) and press Enter: the image shows Making change… straight away, while it is made. It is made from exactly what the image shows at that moment, with its layers as you see them. When it’s ready, the change is a new layer on top, named after what you asked (“Edit – make it night”; see Layers): the editor shows it, and so does the Canvas, where the image was. Nothing to accept, and nothing is lost: everything underneath stays as it was, so hiding the edit’s layer shows the image as it was before. The next change is made from what the image shows then: hide or move layers first, and that is what it starts from. One change at a time. If you leave the editor while a change is being made, it carries on, and the image shows it when it’s ready. When you close the editor, the image on the Canvas shows what the editor showed straight away, while it is saved as the image’s picture. Open the image again and its layers are as you left them. The image itself is never shown above the prompt; images you chose with Use as reference (or add with +) are, to guide the change, and they go with it. Changes use Nano Banana 2.1 unless you choose another model (for the most precise edits, try GPT Image 2.5 Sunburst, or Flare for faster ones), and each one is charged as it is made, even if you hide its layer later.

A change keeps the image’s resolution: an image about 1K wide (up to 1,280 pixels on its longer side) is changed at 1K, one up to about 2K (2,560 pixels) at 2K, and a larger one at 4K. A 2K image stays 2K however many times you change it. The price beside the send button is for that size ($0.045 at 1K, $0.068 at 2K, $0.155 at 4K with Nano Banana 2.1) before you send.

Zooming and moving around

To zoom, hold Ctrl (Cmd on a Mac) and scroll over the image, or pinch on a trackpad. It zooms about the pointer, so what you point at stays under it. Scrolling on its own moves the view. You can also drag with Space held, or with the middle mouse button. Ctrl+= and Ctrl+- zoom in and out, Ctrl+0 fits the image to the window, and Ctrl+1 shows it at its own size. The zoom control at the bottom left does the same: − and + step, and the percentage (100% is the image’s own pixels) fits it again. Every tool works zoomed: paint fine detail close up, or zoom out to drag a crop well past the image’s edges. Cropping, extending or upscaling fits the image again.

On a touch screen, two fingers move the view: drag them to pan, and pinch to zoom around the point between them. One finger works the tool in hand: it paints with the Brush, and drags the crop frame, a layer’s handles or a target’s edges, which a fingertip can take from a little further away. A pinch that starts on the image with the Brush in hand paints nothing. An Apple Pencil works like a mouse.

On a phone the image fills the width, and the panels keep out of its way: Layers opens the Layers panel in a sheet from the bottom, and a tool’s settings fold to their name at the top; tap the name to open them, and again to close them.

Changing chosen things

To change particular things in the image, choose Region (the dashed square) in the editor’s toolbar and drag a rectangle around each thing you want to change: a jewel, a lamp, both earrings. You can draw up to eight. Select one to move it, drag its handles to resize it, or remove it with its ×, or with Delete. With more than one, they’re numbered. The prompt then reads “Describe what to change in this area…” (or “…in these areas…”), and one prompt changes every thing you’ve chosen: “make it a dog’s head”, “make them flowers”.

The rectangles point at what to change; they aren’t a frame the change has to stay inside. A dog’s head that’s naturally larger than the jewel it replaces is kept whole, with its shadow and soft edges, and the rest of the image stays as it was. Anything in the image clearly unrelated to the things you chose is restored to exactly what it was. The result is a new layer on top (“Region – make them flowers”): the model’s whole picture, with a mask that shows only the new thing, so everywhere else the image beneath shows through exactly. Hide the layer and the image is as it was. If the mask misses part of the new thing, or leaves an edge, paint it with the Brush (see below). The rectangles are cleared once it’s made.

Press Esc to remove the selected (or last) rectangle; with none left, Esc puts Region down, and pressing it again leaves the editor.

Layers and removing a background

An image in the editor is made of layers, shown in the Layers panel at the bottom right, top first; the image sits beside the panels, never under them. An image you haven’t layered is one layer, the image itself. Changes show straight away, and you can keep changing layers while a change is being made: what it makes is added, and your choices stay as you made them.

+ adds a New layer (blank and transparent, named “Layer 1”, “Layer 2”…), a Group or an adjustment layer, above the layer you’ve chosen.
The eye shows or hides a layer. Press an eye and drag up or down over the others to show or hide them all the same way, as in Photoshop; Ctrl+Z undoes the whole sweep at once. Double-click a name to rename it.
Drag a layer up or down to reorder it, onto a group to put it in, or out of a group to take it out. With several chosen, dragging any of them moves them all together, in their order. A group’s arrow collapses or expands it; hiding, locking, blending or fading a group does the same to everything in it.
Shift-click chooses every layer from the chosen one to the one clicked; Ctrl-click (Cmd-click on a Mac) adds a layer to those chosen, or takes it away. Group selected (or Ctrl+G) puts them in a new group where the topmost was.
Choose a layer and its blend mode (Normal, Multiply, Screen, Overlay, Soft Light, Hard Light, Colour Dodge, Colour Burn, Darken, Lighten, Difference, Hue, Saturation, Colour, Luminosity) and opacity show above the list.
Alt-click a layer to clip it to the layer beneath: it shows only where that layer does. Several layers can clip to one.
Right-click a layer, or use ⋯, to rename, duplicate, add or delete its mask, clip or release it, lock or unlock it, group or ungroup it, or delete it. A locked layer can’t be painted, moved or deleted. An image always keeps at least one layer.

Click a layer to choose it for the tools; click it again to choose none. A layer made by a Region edit is marked Change: it holds only what changed, and means something only over what is beneath it.

With a whole image layer chosen, Region changes that layer alone, and the result goes directly above it.
With no layer chosen, Region and the prompt change the image as it shows.

To remove a background, choose the layer and choose Remove background in the editor’s toolbar (with only one layer, it’s chosen for you). A Change layer has no background of its own, so Remove background isn’t available for it. The image shows Removing background…, then the layer shows only its subject: it now has a mask, shown beside its picture in the Layers panel. Nothing is cut out of the picture: paint the mask to bring back what it hides, or choose Delete mask to have the whole layer again. Where the image is transparent the editor shows a light checkerboard, and on the Canvas you see what’s behind it.

Moving, scaling and turning layers

Choose Select in the editor’s toolbar (or press V). The chosen layer gets a box around what of it shows: for a Region’s change, just the new thing, not the whole picture.

Drag inside the box to move the layer; the arrow keys move it a pixel at a time (with Shift, ten).
Drag a corner to scale it; its proportions stay the same. Hold Shift to scale it freely, wider and taller on their own.
Drag the middle of a side to stretch it one way only.
Drag the round handle above the box to turn it; hold Shift to turn in 15° steps.
Click another layer’s pixels to choose that layer instead. Press Enter, or click where nothing shows, to choose no layer.
With a group chosen, everything in it moves together.

Scale and Rotation at the top right set them exactly; double-click Rotation to set it back to 0°. Each change shows as you make it and is kept once you let go. A locked layer can’t be moved. Press Esc to put Select down.

Adjustment layers

An adjustment layer changes the colour and tone of everything beneath it, without changing any picture: hide it, fade it, mask it or delete it, and what is beneath is as it was. Choose + → Adjustments, Hue/Saturation or Curves; it goes above the layer you’ve chosen, and its settings appear at the top right whenever it is chosen. With a tool in use, a tab above the settings switches between the tool’s and the adjustment’s. Every change shows as you make it.

Adjustments: exposure, contrast, highlights, shadows, saturation, temperature and tint, each from −100 to +100. Double-click a setting’s name to put it back to 0.
Hue/Saturation: hue, saturation and lightness for the whole image (Master) or for one range of colours at a time (reds, yellows, greens, cyans, blues, magentas). A dot marks each range you’ve changed.
Curves: a curve for brightness (RGB), or for red, green or blue alone. Click the curve to add a point and drag it; double-click a point, or drag it off the side, to remove it.

Alt-click an adjustment layer to clip it to the layer beneath: then it changes only that layer. In a group, an adjustment changes everything beneath it, as if the group weren’t there (the group’s blend mode is Pass Through); give the group another blend mode to keep its adjustments to what is in it.

Masks and the Brush

A layer with a mask shows where its mask is white, is hidden where it is black, and shows partly where it is grey; what is beneath shows through at exactly the same edge. Region edits and Remove Background make masks, and Add mask gives any layer or group one (all white, so nothing changes until you paint it).

Choose Brush in the editor’s toolbar (or press B), and its settings appear at the top right: size, softness (at 100% the brush fades all the way from its centre), opacity and flow (below 100%, paint builds up as you go over it). [ and ] make it smaller and larger. Its edge stays smooth at any softness. Hold Shift as you paint to keep a stroke straight across or straight down, or Shift-click to paint a straight line from where your last stroke ended.

Click a layer’s mask (beside its picture in the Layers panel) and the Brush paints the mask in greys: white shows the layer, black hides it, grey shows it partly. The layer’s picture is never changed. Double-click the mask to see it on its own, in black and white, and paint it there; double-click it again (or press Esc) to go back.
Click a layer’s picture and the Brush paints it in colour. What you paint on is kept: the painting becomes the layer’s own picture, and the original stays as it was.
X swaps the colour with the second one (on a mask, white with black).

Each stroke shows as you paint it and is kept a moment after you stop. Press Esc to put the Brush down.

Clone

Choose Clone in the editor’s toolbar (or press S) to paint a layer with its own pixels: to repeat something, or cover a blemish with what is beside it.

Alt-click where to copy from, or choose Pick source (or press S again) and click there.
Paint. A small circle shows where it is copying from as you go.

The distance stays the same for every stroke until you pick again. Sample sets what it copies: This layer, the Layers below it, or All layers as they show. To keep a copy apart, add a new layer, choose Layers below, and clone onto it. Size, softness and opacity are the Brush’s. What you paint is kept like a Brush stroke, and the original stays as it was.

Fill

Choose Fill in the editor’s toolbar (or press G) and click the image: the pixels of the chosen layer connected to where you clicked, and like it, are filled with the colour (on a mask, the grey). Tolerance sets how alike a pixel must be to be filled: at 0% only exactly the same colour, higher to take in more. Opacity sets how much it covers. Fill whole layer (or Fill whole mask) fills all of it. Fill uses the Brush’s colour, and X swaps it. What you fill is kept like a Brush stroke: the original picture stays as it was.

Relight

Relight in the editor’s toolbar changes the lighting of the image as it shows, and nothing else, as a new layer above it. Start from a preset (Soft daylight, Golden hour, Hard noon sun, Beauty key + fill, Dramatic rim, Product studio) or set it up yourself:

Natural: drag the sun across the sky. The middle of the dome is overhead and its edge the horizon, with the camera at the bottom. Then set its intensity, its softness (crisp shadows to overcast) and its colour temperature (warm candlelight to cool shade).
Studio: the subject seen from above, with the camera below it. Drag lights to place them, + Light to add one (up to five), and choose a light to set its role (key, fill, rim or accent), intensity, softness and colour.

Choose Relight; the button shows its price and the model that makes it: Nano Banana 2.1, or Nano Banana Pro (slower and dearer, and the most faithful to everything but the light). The layer is named after the preset or the setup (“Relight – Golden hour”). Hide it to see the image’s own light again.

Cropping and extending

Choose Crop in the editor’s toolbar (or press C). A frame appears over the image:

Drag inside it to move it, or drag its edges and corners to size it.
Drag past the image’s edges to extend it: the new room is transparent.
Choose a ratio to keep: Original, 1:1, 4:5, 3:2, 4:3, 16:9 or 9:16 (Free keeps none).

Choose Apply (or press Enter) and the image becomes the frame. Nothing is cut: what falls outside stays in the layers, and cropping out again brings it back.

Fill new areas fills the room past the image’s edges by continuing the scene, with Bria Expand, which keeps the image itself exactly as it is. Choose it with the frame dragged out (it applies the frame first) or after extending. It goes on a layer beneath everything, so it shows only where nothing else does, and the button shows its price. Press Esc to put Crop down without applying it.

Split layers

Split layers in the editor’s toolbar takes an image apart into layers. It has two choices.

Quick (free) splits the chosen layer (with only one, it’s chosen for you). The layer becomes a group named after it (“Original – split”) holding:

its Subject;
its Background;
the layer as it was, hidden.

Subject and Background are the same picture, each with its own mask, and they meet exactly. Show, hide, move, mask, adjust or delete each on its own. Nothing is copied or lost: delete the group’s parts and show the original to have it as it was. Like Remove Background, it works on a whole picture, not on a Region’s change.

High quality takes apart the whole image as it shows, with Seedream 5 Flash Layerize. A group (“Layers – split”) arrives on top holding:

the scene’s Background, with what was in front of it filled in;
each thing in it (up to 16) as a layer of its own, named “Layer 1”, “Layer 2”… and placed exactly where it was.

Together they look like the image did, and everything beneath is left as it was. The button shows the price of each layer: you are charged only for the layers it makes, the Background among them, so an image taken into a Background and three layers costs four layers’ price.

Upscale

Upscale in the editor’s toolbar makes the image as it shows again larger, with finer detail, as a layer on top. Topaz Wonder 3.5 makes it 2× or 4× its size; Recraft Crisp, much cheaper, enlarges it by its own amount. Each button shows the new size and its price. The image grows with it: every other layer is enlarged in its place, so nothing moves, and transparency is kept.

Animate

Animate in the editor’s toolbar makes a video from the image as it shows. Describe how it moves (“the waves roll in, the camera drifts slowly right”), choose a model, and choose Animate. Each model shows its price, and the first, Seedance 2.0 Fast, is the recommended one (Auto). Seedance 2.0, Wan 3, MiniMax H3 Turbo and Max, and Kling 3 and 3 Pro are there too.

The video is new, a 5-second clip starting from the image: the image itself doesn’t change. While it is made, Animating… shows on the image. On a Canvas the video appears beside the image. Once it is made it opens in the video editor (see Making videos). It is charged as it is made, like every model.

Undo and redo

Ctrl+Z (Cmd+Z on a Mac) undoes the last change in the editor: a layer shown, moved, grouped, blended or deleted, a Brush stroke, a fill, a crop, a new layer, a split, a removed background, or a change’s layer arriving. Ctrl+Shift+Z (or Ctrl+Y) redoes it. On a touch screen, tap with two fingers to undo and with three to redo. A run of quick changes to the same thing (a slider dragged, a layer nudged) is one step. Upscaling can’t be undone: hide its layer instead.

Versions

The versions button at the top of the editor names what the image is now: a version (“Original”, “Version 2”), or Current, your latest edits. It lists the image’s saved versions, newest first, each with its number, and Current above them. An image starts with one version, Original (and, if it was changed before layers, one for each change it had). Only Create version makes a version: keeping the image exactly as it is now, every layer as it is. Changing layers and making changes never do; the image is then Current. Choose a version to restore it: the image is that version again, and your edits are set aside as Current, so choosing Current brings them back. If you change the image while it’s an older version, those changes are your latest edits: Current from then on. Anything made since a version stays in the layers panel, hidden, and nothing is removed. What the image is now is ticked: only ever one.

What the layers show is what the image shows everywhere: on the Canvas, when you download it, and when you use it as a reference. A moment after you stop changing layers (and when you leave the editor), the Canvas catches up, showing exactly what the editor shows: blend modes, groups and masks included. Showing, hiding, moving, renaming and changing layers never makes a new version of the image, and nothing piles up: only the picture the layers show now is kept.

To go back, choose Close (×) at the top right or Done (✓) in the editor’s toolbar, press Esc (outside the prompt, once you’ve put Region down), or use your browser’s Back. The image returns to its place, and the Canvas is exactly as you left it: the same view, with the image still selected.

Paste (Ctrl/⌘ + V) an image copied from another app or a browser while the editor is open, and it arrives as a new layer above the chosen one, at its own size in the middle of the image (smaller only if it wouldn’t fit). It never paints over a layer, and Ctrl/⌘ + Z takes it back.

Add mask shows the new mask beside its layer at once, and chooses it, so the Brush paints the mask straight away.

The image’s address is the same wherever you opened it from. Opened on its own (pasted into a new tab, say), it shows the same editor without the Canvas; leaving it then takes you to the Space (or Drafts) the image belongs to. Only people who can open that Space can open the image this way.