Manual 06 layers selections and guides - orlandomenke/Lightbox GitHub Wiki
Layers, selections and guides
Layers
Folders, blend modes, per-layer opacity, visibility, lock and alpha lock. Thumbnails show what is actually on the layer.
A thumbnail is a preview, and is drawn at preview size. It is rendered small rather than rendered full-size and shrunk β which matters only because Lightbox decides the fine grain of a mark from where the mark is, so a scatter or a grain in a 32-pixel thumbnail will not land on exactly the pixels it lands on at full size. Nothing about your artwork is affected, and it is the reason the timeline and the layer docker cost nothing to keep up to date: making them faithful to the pixel would mean rendering the whole document once per drawing, which is what used to make the canvas pause every time you lifted the pen.
There is one kind of layer, and it holds everything. Strokes, imported pixels, symbol placements β a layer is not committed to one of them. Lightbox used to ask you to pick Raster or Vector when you added a layer, and marked each row R or V; the choice never changed what you could draw, because both kinds took the same tools and made the same marks. All it did was decide what a layer would later refuse. Documents saved before this still remember which they were, so nothing about your files changed β the question is simply gone.
A new document opens with a locked Background layer holding the paper, and a paintable layer above it. On a transparent document there is no paper layer β just an ordinary unlocked layer.
Opening a file selects a layer you can draw on β not merely the first one above the paper. A layer that is locked, hidden, or inside a locked folder is skipped, so the first stroke after opening lands instead of going nowhere. If every layer is out of reach the selection still goes to a real layer rather than the paper, and the status line says which lock is in the way.
A locked row shows a shut padlock and a quieter name, and nothing round the row: the highlighted row is always the one you are drawing on, never a locked one. A layer or folder inside a locked folder shows the padlock shut as well, dimmer, because the lock that is set is the folder's β unlock the folder and those rows are as they were. (A locked row used to be outlined, which made it look like the active layer.)
And it reopens on the layer you left it on, the way it reopens on the frame you were parked on. A file saved while you were inking comes back on the ink layer. If that layer has since been deleted, locked or hidden, the selection falls back to the first one you can draw on rather than honouring it β a remembered layer is never the reason a stroke goes nowhere.
Ctrl+click a layer thumbnail to select its opaque pixels.
A new layer goes where you are working: directly above the active layer, and into the active layer's folder if it is in one. Click a folder's header first and the new layer goes on top inside that folder instead β an empty folder too. (It used to land on top of the whole stack whatever was picked, outside every folder.)
The everyday layer verbs are on the Layer menu too, addressed to the layer you are on: new layer, new folder, move up and down, merge down, the three checkboxes (visible, locked, lock transparency), select contents, blank content and delete. The docker keeps the things a menu is wrong for β the blend mode and opacity controls, drag-reordering, and the right-click menu's row-targeted extras (linking, exports, the rig).
Folders
A folder is an item in the stack in its own right. It can hold layers, other folders, or nothing at all, and it stays where you put it either way.
- New folder (the folder button under the docker, or Layer β New folder) makes an empty folder directly above what you are on: above the active layer, inside that layer's folder if it is in one, or above the folder you have picked. With nothing to go above, it goes on top of the stack. The folder you are in is never touched.
- Group layers (Ctrl+G, Layer β Group layers, or Group into a new folder on a row's right-click menu) puts the selected layers β and a picked folder β into a new folder where the topmost of them was. Layers that were apart are gathered together, which changes the stacking; undo puts them back. This is Photoshop's Group from Layers and Krita's Quick Group; New folder is their New Group.
- Folders go inside folders, as deep as you like. Each level is indented one step, and hiding or locking a folder hides or locks everything in it, at every depth.
- A hidden folder shows on every row it reaches. Each layer and folder inside it is dimmed and its eye is drawn shut, at every depth, so the docker never shows an open eye on something the canvas is not drawing. Those rows' own eyes are not changed by it: show the folder again and everything comes back as it was, including the layers you had hidden one by one. The X-sheet's rows do the same.
- A folder's colour washes its whole row, and the rows of the layers inside it more faintly, so a folder reads as one band down the docker rather than as a mark beside its header. Choose it from Folder color on the header's right-click menu. A folder inside another shows its parent's colour until you give it one of its own, and then everything inside it follows that instead; Same as parent, on the same menu, gives the colour up again. The row you are drawing on and the rows you have selected are still the brightest things in the list. A file saved before this recorded a colour on every folder, so its subfolders keep theirs until you choose Same as parent.
- Click a folder's header to pick it on its own β an empty one included. With a folder picked, a new layer or folder goes inside it, and Delete deletes the folder.
- Deleting a folder deletes everything in it: its layers and the folders inside it, as one undo step. To keep what is inside, use Ungroup on the header's right-click menu or its β button instead: the folder goes and its contents stay where they were, one level up. A folder with a locked layer or locked folder inside it is not deleted until that is unlocked.
A file made before folders could nest opens exactly as it was. If a file has a folder whose layers are not next to each other β something older versions could leave behind β its header simply appears again above the second part, rather than the layers being reordered on opening.
Finding a layer from the canvas
Ctrl+Shift+click on the drawing makes the layer you clicked on the active
one, with any tool in hand. Its row scrolls into view in the Layers docker and
the status line names it.
- It picks the topmost layer that shows something there: the mark your eye is on. Where two layers overlap, the upper one.
- A hidden layer is looked through. You cannot see it, so you did not click on it. So is a layer turned down to almost nothing, and the background: that is paper, not somewhere a mark lives.
- You do not have to hit a hairline to the pixel. A click up to ten screen pixels off still finds it. Zoomed far out the reach stops at 16 pixels of the drawing, so that a click never finds something a long way off. A click squarely on one layer's mark is never taken by another layer's line passing close by.
- Bare paper changes nothing, and the status line says there was nothing there. Onion-skin ghosts and reference images are not layers, so a click on one finds whatever layer is under it, or nothing.
With the Move tool, or inside a selection, the same keys already drag:
Ctrl moves the whole layer or what the selection holds, and Shift keeps it
to one axis. So there a click finds the layer and a drag moves, the way it
always did. A press that travels less than three screen pixels is a click, and
moves nothing.
Where it does not reach: on a guide while the Move tool or the Arrow has it, on a handle (the camera's, the symmetry axis's), and while the canvas is given over to something else β a transform, the reference grid, the rig.
It selects that layer alone, the same as clicking its row, and it is looking rather than editing: nothing goes into the history. A layer you find this way may be locked; the pointer will say so when you go to draw on it.
Working on several layers at once
Click a row to make that layer active. Ctrl+click another row to add it to the selection β or Ctrl+click a selected one to drop it. Shift+click takes every row between the last one you picked and this one; Shift+click again somewhere else re-ranges from the same starting row rather than adding a second run, so overshooting is corrected with one more click.
Ctrl and Shift work anywhere on the row β on the eye, the locks, the arrows or the name alike. With either key held, a click on a row's button selects the row instead of pressing the button. The one exception is the thumbnail, as in Photoshop: Ctrl+click a layer's thumbnail to make its drawing a selection on the canvas, Ctrl+Shift+click another thumbnail to add that drawing to it, and Ctrl+Alt+click to take one away.
The X-sheet picks layers the same way. Ctrl+click or Shift+click a layer's name on the X-sheet to add it or take the range. It is the same selection the docker keeps, so a set picked in one place is still picked in the other, and a plain click on either goes back to one layer.
Clicking a folder's header picks the folder: its header lights up, every layer in it (at any depth) is selected, and its top layer becomes the active one. An empty folder is picked all the same, with no layer selected. Ctrl+click another header to pick that folder too, so every picked folder lights up, and Ctrl+click a picked one to take it back out. A folder whose layers you had all picked one by one counts as picked. Shift+click a header to range up to it. Ctrl+clicking a layer out of a picked folder unpicks the folder, since it is no longer whole.
With folders picked, Delete removes the folders with everything in them, along with any other picked layers, as one undo step. The eye and the lock on any picked folder cover every picked folder.
Every selected row β and every picked folder β has a light bar down its left edge and a tint, and one of them β the last you clicked β is tinted more strongly. That one is the active layer, and it is where the next brush stroke lands. There is always exactly one, which is why Ctrl+clicking the only selected row does not deselect it: the app would have nowhere to paint.
Anything you then do to a row in the selection is done to all of them, as one undo step:
- delete, and blank the content
- move up and move down β the block keeps its order and stops as one against the top or bottom of the stack
- the eye, the lock and the alpha lock
- In exports β never, always, or leave it to the export
- Group into a new folder (Ctrl+G), which puts the whole selection in one folder
Anything you do to a row that is not in the selection is done to that row alone. Right-clicking a layer you have not selected is not a trap.
Copying a layer or a folder
Copy layer and Paste layer are on the row's right-click menu in the docker and on a layer's name on the X-sheet, and Ctrl+C / Ctrl+V do the same when a layer row or a layer name on the X-sheet has keyboard focus. Right-clicking a layer that is already in the selection copies the whole selection.
A paste is a new layer with everything on it: every drawing and hold, the mask, effects, blend mode and opacity, and the rig it follows. It lands directly above the active layer, in that layer's folder, named Ink copy (Ink copy 2 and so on if that is taken). Several layers paste together in their original order, the pasted layers become the selection, and the whole paste is one undo step. Pasting twice gives two independent layers.
A folder copies whole. Pick a folder by its header and press Ctrl+C, or use Copy folder on the header's right-click menu: the folder comes with everything in it β its layers, the folders inside it, and any empty folder in its place. Ctrl+V pastes it as a new folder, named Hero copy, wherever a new layer would go; Paste into folder on a header pastes at the top inside that folder. Everything inside keeps its arrangement, the pasted folder is picked, and the paste is one undo step. An empty folder copies too.
What a copy does not keep is what tied the original to other layers: it does not join the original's link or fluid group, and it is never the paper.
A layer can only be pasted into the document it was copied from β pasting into another says so and does nothing. With nothing selected, Ctrl+C copies the active layer this way wherever the focus is β see Copying lines out, and pasting them back below. While a layer row or a layer name has focus, Ctrl+C and Ctrl+V mean layers even with a selection up on the canvas. Copy the selected layers and Paste the copied layers are also in Edit β Configure β Shortcuts, unbound, for a key that is always a layer.
Reordering by dragging
Rows also drag: pick a layer up and drop it where it should go. Dropping on the upper half of a row lands above that row, the lower half lands below it. The layer takes the folder of wherever it lands β dropping between a folder's members joins the folder, dropping beside a loose row leaves it β and the whole drop is one undo step. A drag moves the one row you picked up; the β²/βΌ buttons remain the way to move a multi-selection as a block.
Hold the drag near the top or bottom of the docker and the list scrolls towards you, so a layer can be dropped on a row that was not showing. It starts after a moment's pause β a row picked up near the edge does not send the list running β and then keeps scrolling while you hold still β faster the closer to the edge, fastest if you drag past it β and stops when you move back to the middle, reach the end of the list, or let go. The drop line follows the row that is now under the pointer.
Every part of the docker is a place to drop, including the thin gaps between rows and the indent in front of a folder's layers: each belongs to the nearest row. Letting go outside the docker cancels the drag. A pen has to travel a little further than a mouse before a press becomes a drag, so a tap that wobbles stays a tap.
A folder header drags too, and takes the whole folder with it β every layer and folder inside it, in their own order. It lands above or below wherever you drop it, or inside another folder, exactly as a layer does. An empty folder drags the same way. The one drop a folder refuses is into itself or into a folder inside it; nothing is offered there.
A folder header is three drops in one row, the same whether you are carrying a layer or a folder, and whether the folder is open or collapsed:
| Where on the header | What happens |
|---|---|
| Top third | Lands just above the folder, beside it |
| Middle third | Goes into the folder, at the top |
| Bottom third | Lands just below the whole folder, beside it |
On an open folder the line for "below" is drawn under the folder's last row, because that is where the drop lands β not under the header, which is where "into" lands. This is how a folder is put between two others: aim at the top or bottom third of the neighbour's header. And when a folder is the very last thing in the docker β a transparent document has no paper under it β the bottom quarter of its last layer also takes a layer out below the folder.
A drop that would leave everything where it is does nothing at all: it is not an undo step and does not mark the document changed.
While you drag, two things tell you what will happen: a small label under the pointer naming what you picked up, and a line on the row you are over showing which side it will land on. A folder about to swallow what you are carrying lights up instead of drawing a line, because "inside" is not an edge. If neither appears, the drop will not do anything β dragging a folder onto one of its own rows, for instance.
The onion-skin toggle (β) is the exception: it is always one layer, even with several selected. The eye and the locks describe the drawing, and picking five layers means all five. Onion skin describes what you are looking through while you work on one of them, and which layers are in the stack is something you tune as you go β the background off, the layer under your hand on. Sweeping it across a selection would clear that arrangement and give nothing back.
The keyboard and the shortcut bar say "active layer" and mean it: Toggle lock, Toggle alpha lock and the bar's β act on the active layer alone, never on the selection. The docker's own rows are the selection-aware ones.
Clicking a cel in the timeline, walking the stack with the arrow keys, or opening a document all set the active layer the ordinary way, and that starts the selection again from the one layer.
Linking layers β one drawing across several
A character is rarely one layer. Lines, colour, details, effects: four layers that are really one drawing. Linking says so once, and what the link carries then reaches all of them β on every frame, including the ones you have not drawn yet.
- Ctrl+Shift+right-click links a layer to the one above it.
- Ctrl+Alt+right-click links it to the one below.
- Shift+right-click takes it back out.
- Right-click and open Linking for the same options by name, plus the switches below.
Linking to a layer that is already in a link joins that link rather than starting a second one, so a four-layer character is three gestures and ends with one link.
Linked rows are indented and carry a bracket down their left edge, so a set reads as one at a glance.
What travels a link is up to you, per property. A link can carry several things, one, or nothing at all:
- Share the rig β every layer follows the bone any one of them names. On by default when you make a link, because rigging is usually why you made it.
- Share alpha lock
- Share visibility β hide the lines and the colour goes with them.
Nothing else travels. A link made to rig a character does not quietly start sharing anything you did not ask it to.
The gestures read the neighbour once, at the moment you make them. After that the layers know each other by name, so reordering the stack afterwards never breaks a link or attaches it to something else.
Making a layer follow the rig
Right-click a layer and open Follows the rig:
- The selected bone β every stroke on the layer follows the bone picked in the Bone tool.
- The whole skeleton β weighted by joint, the way auto-bind does it: each part of a drawing follows the bone it is on, blending only near the joints. A layer bound this way before joint weighting existed keeps its softer weighting until you choose this again.
- Nothing β off the rig.
This is per layer, so it covers drawings you have not made yet and frames you have not reached. Scrub the timeline and every drawing on the layer follows the pose, in playback and in every export. Painted weights on an individual line still win over it β that is what the weight brush is for.
Your lines are never changed by this. A rigged layer decides where marks are drawn; the drawing stays exactly as you made it, so unrigging the layer returns it untouched. Baking a drawing freezes the pose into it and leaves the layer rigged, so the drawings you make afterwards still follow.
Drawing on a posed layer puts the mark where your pen was. The layer may be standing in a pose when you draw on it; the new line lands under the pen and stays there, and it follows the rig from then on like everything else on the layer. Under the hood the line is stored in the rest pose so the rig can move it β you never see that. On a layer bound to the whole skeleton, a line you draw keeps the weights it was drawn with, measured against where the bones stood at the time, so it does not re-weight itself when the pose changes. One consequence worth knowing: if you later take that layer off the rig, lines carrying their own weights keep following it while the rest of the layer stops β unbind them too (or bake first) if you want the whole drawing to let go. Shapes, fills, gradients and pen paths land where you made them in the same way; a pen path keeps its points and loses its curve handles in the process.
A layer's own choice beats its link's, so you can rig the effects layer to a different bone from the lines it is linked to.
Merging a layer down
Ctrl+E merges the active layer into the one below it β Photoshop's and Krita's key. The row's right-click menu has it too (Merge down), for merging a layer that is not the active one. The merged layer keeps the lower layer's name, opacity, blend mode and folder, and the merge is one undo step.
It walks the exposure sheet drawing by drawing: where both layers hold, the merged layer holds, so animating on 2s survives a merge. Where either layer keys a new drawing, the merged layer gets one combined drawing.
Your strokes survive a merge. Two plain drawings merge by keeping both stroke records as one list, still editable line by line. Where that alone would change the picture, the drawing being merged keeps its strokes together as a group inside the merged drawing, and the group carries what its layer did:
- a blend mode or opacity on the layer being merged;
- a mask on it, a clip onto the layer below, or layer effects;
- erasers or clear-regions on it β as loose strokes they would start erasing the lower layer's marks, which they never touched. Here you have a choice β see below;
- alpha-locked, smudge, blur or colour-mixing strokes, and imported pixels β all of them read what is underneath, and "underneath" changes when the layers become one.
A group is drawn the way its layer was: its strokes on their own, then put onto the drawing through the blend, opacity, mask and effects it carries. So the merge looks exactly as the two layers did, nothing has been turned into a picture, and the AI inbetweener still reads every line. A mask or effects on the lower layer move into a group around that layer's drawings the same way, and come off the layer β left on it, they would start applying to the drawing that was merged in.
The decision is per drawing, not per layer: one blended drawing does not put the other eleven in groups. As in Photoshop, a blend mode lands against the layer directly below β if it was interacting with layers deeper in the stack, that part of its look can shift.
What a group can do. It moves, scales, rotates and resizes with its drawing. Under a selection, the part of it inside the selection moves and the rest stays, cut at the selection's edge. Whatever you draw or erase afterwards lands on top of it, so erasing on a merged drawing erases all of it. Onion skin, motion trails and the inbetweener all see the lines inside it β except a group that shows nothing at all, at zero opacity or behind an empty mask, which the inbetweener leaves out. A line that a mask only partly hides is still read whole.
Clicking a merged line selects the whole group. It highlights as one thing β the status bar says 1 group selected β and Delete, copy, the arrow keys and a drag move or remove it whole. Recolouring waits: a group's colours are its lines', and changing those is editing inside it.
Split group puts a merge back. Layer βΈ Split group, or the layer row's right-click menu, takes the drawing that was merged in back out of every cel of the layer, onto a new layer directly above, named as it was and with the blend mode, opacity, clip, mask and effects it had. It works on the group you have selected, or the topmost one in the drawing at the playhead, and it is one undo step. Before you draw anything more, the two layers look exactly as they did before the merge β which is also how the merged one looked, unless the merged layer itself has a blend mode, an opacity below 100% or a clip. A group that is the merged layer's own drawing, kept with its mask or effects, has no other layer to go back to and is left alone. Anything you drew on the merged layer afterwards stays there: an eraser stroke that reached the group is copied onto the split layer too, so what you rubbed out stays rubbed out, but a mark you drew over the group now sits under it β the status bar says how many. It has no key by default; give it one in Edit βΈ Configure βΈ Shortcuts.
Planned: picking, reshaping or recolouring a single line inside a group. Until then, split the group to work on its lines, or choose Cut the lines below when you merge, which leaves every line loose.
A document that holds a group needs a Lightbox that knows groups. A version from before groups will refuse to open it rather than open it with the merged drawing missing. Its message is blunt β could not be converted to ToolKind β because those versions never learned to say "saved by a newer Lightbox" for this; from this version on, a file too new to read says so in those words. Documents with no group in them are saved exactly as before and open anywhere.
Erasing: cut the lines, or keep the erasers. A line drawing nearly always has erasing in it. When erasing is the only thing standing between a drawing and plain strokes, Lightbox asks which you want:
- Cut the lines. Each line is cut where you erased it and the erasers are dropped, the way an eraser works in a vector program. The drawing becomes plain strokes: every line can still be picked and reshaped on its own. It is nearly exact rather than exact. A soft or half-strength eraser becomes a clean cut. Each piece stops where its own round end would first touch what you erased, so ink does not come back into the gap, and the corners of a cut come out slightly short instead. An eraser that only shaved a line's edge without reaching its middle leaves the line whole, and so does the stretch where an eraser ran along a line after crossing it. A textured brush starts its grain afresh at a cut. On thin line art the difference is hard to find β measured on three real merges it was 0.00%, 0.11% and 0.27% of the painted pixels β and it grows with the width of the brush: 2.7% on a deliberately hard case, a 24-pixel line under a 30-pixel eraser.
- Keep exact. Exactly what the two layers looked like: the drawing keeps its strokes and its erasers together as a group.
Tick Remember this choice and it stops asking; Edit βΈ Configure βΈ Drawing shows the remembered answer and "Ask every time" takes it back. With a remembered answer, Ctrl+E simply uses it; with none, it keeps the erasers.
A drawing goes into a group whatever you choose when something else needs one β a blend mode, opacity, a clip, an alpha-locked or smudge stroke β or when cutting would not give the picture you had:
- the erasing reaches something that is not a plain painted line β a fill, a gradient, text, a blur or a wet-media stroke;
- you erased inside a selection, which confined the eraser to it;
- the eraser, or a line it might have crossed, was drawn with symmetry or tile wrap, so its mirrored copies matter too;
- the drawing already holds a group from an earlier merge.
Those drawings keep their erasers, and the others on the same layer are still cut.
The few drawings a merge still turns into pixels are the ones shaped by something that is not theirs to carry:
- a drawing kept inside a folder's shape β the shape is the other layers in the folder;
- a drawing with a placed symbol on it, when it would otherwise need a group β a symbol is placed on a drawing, not in its stroke list, so it would slip out from under the blend or mask;
- a layer clipped onto a layer that has effects β the clip follows the lower layer's drawing, and inside a merged drawing it would follow the blurred or glowing result instead;
- a smudge or blur set to read all layers, live β "all layers" included the one it is being merged into;
- a drawing already merged eight groups deep β groups nest each time you merge something that needs one, and past eight the file could no longer be saved, so that drawing flattens instead.
When AI assistance is on and a merge would do that to any drawing, Lightbox asks first, because a flattened drawing is one the inbetweener can no longer read.
Tracing pixels back into strokes
A layer of pixels β a drawing an older merge flattened, an imported image, a scan of something drawn on paper β can be traced back into strokes. Layer βΈ Trace to strokesβ¦ opens a window with the drawing at the playhead traced in front of you, over a faint copy of its pixels, and three controls:
- Read ink from β transparency for a drawn layer, where the ink is on nothing; darkness for a scan or a photo, where every pixel is opaque and the ink is what is dark. Left on the first choice, Lightbox picks by looking.
- Faintest ink β how faint a pixel may be and still count as ink. Lower takes pale pencil lines in; higher keeps only the darkest. On a scan this is the contrast control.
- Shortest line β marks shorter than this many pixels are dropped as specks of noise. A dot is always kept.
Each has a slider to explore by feel and a field to type into.
Trace traces every drawing on the layer that holds pixels onto a new layer directly above, named after it with (traced), with the same blend, opacity, clip, mask, effects and folder, and the same timing. The pixel layer is hidden, not deleted, so you can compare the two and go back. Strokes a drawing already had are copied across on top of the traced ones. One undo step. It has no key by default; give it one in Edit βΈ Configure βΈ Shortcuts.
Each line becomes one brush stroke down its middle, with pressure read from how wide the line is there, so a line that swells and tapers traces as one that does. A crossing comes back as two lines, a T as its bar and its stem, a closed loop as one stroke. The strokes are ordinary strokes: reshape them, and the AI inbetweener can match them.
What a trace cannot give back, because it is not in the pixels: the exact edge of a line comes back as a round brush's edge; the order the lines were drawn in is guessed (top to bottom); the pen's real pressure is guessed from the width; and where many lines tangle, which way each one went is a guess too. Lines that run very close together can merge into one or break into pieces β the faintest-ink slider is the first thing to try.
Planned: tracing colour areas into fills. Tracing reads line art only today.
Masking a layer
A layer mask hides part of a layer without touching its drawing. In Lightbox a mask is not a special grayscale image β it is an ordinary drawing, painted with your own brushes: where the mask has paint, the layer shows; where it has none, the layer is hidden. Erasing on the mask hides again. There is no white-reveals, black-conceals convention to remember.
Right-click a layer and open Mask:
- Add β painting hides. The layer stays fully visible and what you paint on the mask conceals it. This is the vignette workflow: soften an edge, knock a hole in a background.
- Add β painting shows. The layer starts fully hidden and your marks reveal it β for building a reveal by hand.
The two are the same mask one Invert apart, and inverting is a flag, not a repaint. Adding a mask drops you straight into painting it: a small chip appears beside the layer's thumbnail, outlined while your strokes are landing on the mask. Click the chip (or use Mask β Paint the mask) to switch between painting the mask and painting the drawing β while you paint a mask, the canvas shows exactly what your mark reveals or hides, live, and every mask stroke is a normal undo step. Disable switches the mask off without losing it (the chip goes hollow); Delete removes it entirely.
The mask is one drawing that holds across the whole timeline β it does not animate per frame. For an animated matte, use a clipping mask instead: the matte is an ordinary layer, and ordinary layers animate.
Blur and smudge do not work on a mask, because a mask is coverage rather than pixels to rework; paint or erase it instead. A soft-edged brush gives a soft mask edge, exactly as it gives a soft mark.
Clipping a layer to the one below
Ctrl+Alt+G (or right-click β Clip to layer below) clips the active layer: it composites only where the layer below has content β Photoshop's clipping mask, and the standard way to shade or texture a flat colour without leaving its silhouette. A clipped layer's row shows an accent bar, and several clipped layers in a row all clip to the first unclipped layer beneath them, so lines / colour / shading stacks work as they do elsewhere.
A clip stays inside its folder. The base is always a layer in the same folder, so a clipped layer at the bottom of a folder shows unclipped rather than clipping to something outside it, and a layer just above a folder does not clip into it.
Because the base layer is an ordinary layer, an animated matte is just a clipping mask over an animated base: animate the base's cels and everything clipped to it follows, with holds and exposure exactly as the sheet says. If the base has a mask of its own, the clipped layers respect it too. A clipped layer with nothing below it, or whose base shows nothing on this frame, simply shows nothing there.
What clips to a layer follows the pen on it. Erase the base and the clipped layers go with it as you erase; paint on the base and they cover the new paint as it goes down. The same is true of smudging or wetting the base, of painting its mask, and of a layer that is part of a folder's shape. Nothing changes when you lift the pen.
Keeping a folder's layers inside a shape
This is the colouring setup: flats at the bottom of a folder, and shading, highlights and texture above them that can never go outside the lines. Krita calls it inherit alpha. Here it is set once for the folder rather than on every layer.
Right-click the top flat and choose Keep layers above inside this (Ctrl+Alt+Shift+G for the active layer). From then on, every layer above it in the same folder shows only where that layer or any layer under it in the folder has content. One shading layer over separate skin, hair and clothes flats stays inside all three, and a layer you add above later is kept inside too, with nothing to switch on.
- The docker shows it. An amber line runs down the left of every layer that is kept inside and ends on a dot beside the shape layer. Hover the line to see which layer and folder it means.
- Nothing is cut. Strokes that run outside are hidden, not erased. Stop keeping layers inside (the same menu, on the shape layer) shows them all again, and every change is an undo step.
- It stays in its folder. Layers outside the folder are never affected. It needs a folder: on a layer that is in none, the command tells you to put it in one first rather than doing nothing.
- When something shows nothing, the row says why. A ! appears on a layer kept inside a shape whose layers are all hidden, and on a clipped layer with nothing below it in its folder. Hover it for the reason.
- Masks count. A flat's own mask trims its part of the shape, exactly as it trims the flat.
- Merging keeps the look. Merging a kept-inside layer down into the shape layer bakes the trim into the merged drawing. Merging the shape layer down moves the shape to the layer it merged into, unless that layer is outside the folder: then the folder keeps nothing inside any more (one undo brings it back). Dragging the shape layer out of the folder, or deleting it, releases the folder the same way.
Clipping and this work together: a layer can be clipped to the one below and kept inside its folder's shape, and shows only where both allow.
Both masks and clipping reach every render the same way: the canvas, exports, thumbnails and the AI's view of the document all show the carved result.
Adjustment layers and effects
Open the Effects panel (Window β Effects) to filter without repainting. Three filters ship β Levels, Hue / Saturation and Gaussian blur β alongside five layer styles β Drop shadow, Outer glow, Inner glow, Stroke and Bevel β and every one is a setting, never baked pixels: your strokes are untouched, sliders re-render, and undo steps back through every change.
An effect can sit in two places, chosen at the top of the panel:
- Layer β the stack on the active layer, applied to that layer's own drawing before its blend mode and opacity. Blur the shadow layer, grade the ink.
- Scene β one grade over the whole composite, everything the camera sees.
Hue / Saturation is the one effect the Layer scope does not offer: its true colour-wheel maths runs on the composite path, so on a single layer's own drawing it would do nothing. Add it as an adjustment layer and β if you want it on one layer only β clip it to the layer below; that is the same result, and the panel's add row keeps the distinction honest by only listing what will actually change pixels where you are.
One honest caveat: Hue / Saturation's Lightness lifts and lowers everything equally, black ink included β a positive nudge greys linework. For real value work reach for Levels (its gamma and output sliders are the tool built for it); Lightness is for a quick wash.
Layer styles
The styles are the reverse case: a glow, a stroke, a shadow or a bevel decorates a layer's own silhouette, so they live on the Layer scope only β the scene grade and adjustment layers, which work on the composite beneath them, have no silhouette to read and do not offer them. Each style's colour sits in the panel as a swatch and a hex field.
- Drop shadow β distance, size, angle and opacity; the angle is the light direction, so the default 120Β° throws the shadow down and right.
- Outer glow and Inner glow β size, spread (outer) and opacity.
- Stroke β width, opacity, and a position slider: 0 outlines outside the silhouette, 1 paints the rim inside it, 2 centres the line on the edge.
- Bevel β size, depth, light angle, and a direction slider: 0 shades inside the edge (raised), 1 builds a ridge around it. Highlight and shadow colours are yours to change. This is the smooth bevel only β contour and gloss curves arrive with the timeline's curve editor.
Styles follow what you actually see. Mask half a drawing and the glow hugs the half that is left; a blur on the same layer is trimmed by the mask instead, because a blur is part of the drawing where a glow decorates it. Styles apply after the layer's filters, whatever order the stack lists them in. A live stroke grows its layer's glow as you draw it, and merging a styled layer down bakes the style into the merged pixels β exactly what the canvas showed.
The filters
Six filters on the Colour and Detail shelves do the everyday Photoshop jobs, and every one works wherever you put it β a layer's own stack, an adjustment layer, or the whole scene:
- Sharpen β an unsharp mask. Amount is the strength, Radius the size of detail it works on. The radius starts at 2 pixels because anything smaller has no effect at all.
- Find edges β keeps where the picture changes and drops the flats, drawn as dark lines on white.
- Invert β the negative, and its own undo.
- Threshold β everything to black or white either side of a Level. It reads brightness, so a saturated colour comes out one flat tone rather than breaking into a primary.
- Posterize β flattens to a chosen number of Levels, keeping true black and true white at the ends.
- Gradient map β replaces tone with colour: the darkest parts take the Shadow colour, the lightest the Highlight, and Midpoint decides how much of the picture sits in each. A quick duotone, and a fast way to colour-key a whole shot.
Emboss is Planned β the honest version needs machinery that arrives with a later change, and a half-working one is worse than none.
Effects that move: wiggle, flicker and grain
Three effects vary by frame rather than sitting still, and they are what the panel's Animation shelf is for.
- Wiggle shifts the layer by a random-looking amount that changes every few frames β the boil a drawing gets when it is traced again and again. Amount is how far it can stray, in pixels.
- Flicker dips the layer's strength, never above full β a failing light, a strobe, a shape that stutters in and out.
- Film grain lays a fine noise over the picture. It is a Colour-shelf effect and works on what is beneath it, so put it on the Scene scope (the whole composite, which is where grain belongs) or on an adjustment layer. Size is the grain's own size in pixels.
Hold says how often each one changes, in frames β the same unit the exposure sheet uses. A wiggle on a hold of 2 moves on 2s, alongside a sequence animated on 2s; grain defaults to a hold of 1, because film grain moves every frame. Nothing here is random in the sense that matters: the same frame always looks the same, so scrubbing back and forth, exporting twice, or reopening the file next year all give exactly the picture you approved.
A held drawing still moves. Wiggle and flicker read the playhead, not the drawing, so three frames of one held cel boil for all three β which is the reason to reach for them.
Seed is what makes two layers differ. Every effect starts with its own, so two wiggling layers do not move in lockstep; change it to re-roll a motion you did not like, and leave it alone to keep the one you did.
The quickest way in is the layer row itself: right-click β Layer style and pick a style β it lands on that layer and the Effects panel opens with its sliders in hand. A layer carrying any effects wears a small fx chip beside its thumbnail; click it to jump to the panel.
Everything here is a setting, and everything switches off. Each effect has its own checkbox in the panel, and the whole stack has a master switch β the checkbox at the top of the panel, or Layer style β Disable effects on the row. One click silences every effect on the layer while keeping each one's own on/off and every slider exactly as you tuned them; the fx chip goes hollow, the way a disabled mask's chip does. There is no destructive version of any of this β the only thing that ever bakes an effect into pixels is Merge down, which says so first.
Adjustment layers are the third place, and the one Photoshop users will reach for: right-click a layer β New adjustment layer (or use the buttons in the Effects panel) and a layer with no drawing of its own lands above the active one, applying its effect to everything beneath it. It is an ordinary layer everywhere it counts β mask it to grade part of the picture, clip it to the layer below to grade one silhouette, fade it with its opacity, hide it with its eye, and it holds across the whole timeline like any layer. Delete it and nothing was ever touched.
Effects on layers you paint under update live, exports and the AI's view show exactly what the canvas shows, and a document that never opens the panel never stores or pays for any of it. An effect saved by a newer version of Lightbox than yours shows in the stack marked from a newer build β it is kept exactly as saved and simply does nothing here.
Keyframing effect values over time is Planned: the file format already stores animated parameters, but the editor for placing those keys arrives with the timeline's curve-editing work.
Selections and transforms
Marquee, freehand, polygon, ellipse and magic wand, with Shift to add and Alt to subtract β the crosshair wears a small + or β while the modifier is down, so which way the press will combine is visible before you make it. Grow, shrink and feather. A selection clips painting, and the clip is part of the record, so a reload paints the same shape. The wand shows the region it would take as a faint, still dashed outline before you click β faint and still so it cannot be read as a selection you already made.
The polygon shows its working. The first click puts a small ring where the vertex landed, and from then on a dashed band follows the pointer from the last vertex you placed, so between clicks you can see the segment you are about to commit to. The ring stays on the first vertex because that is the one you double-click back to when you close the shape. Escape abandons a polygon you have started; so does reaching for another selection shape, since none of the others can finish it.
Ctrl+D deselects, whatever tool you are holding. Worth saying plainly because a selection clips painting: if the brush seems to have stopped working, the usual reason is a selection still up somewhere off screen, and Ctrl+D is the answer. Ctrl+A selects the whole canvas β except with one of the arrows in hand, where "all" means the objects on the canvas rather than the canvas.
Delete clears what is inside the selection; Backspace fills it with the background colour. Both leave the outline up, because the next thing you usually do with an emptied region is put something else in it, and both are ordinary undo steps. With no selection up, Delete falls back to deleting whatever lines the Arrow has picked.
A selected guide takes Delete first, ahead of the selection β the thing wearing the selection marks is the guide, so that is what the key acts on.
Every one of those is on the Select menu too, next to Image in the menu bar: All, Deselect, Invert, Grow, Shrink, Delete contents and Fill with background. The keys are faster once you know them; the menu is where you find out they exist, and where you check what a key does without pressing it. Entries that need a region to work on β Grow, Shrink and Fill with background β grey out until there is one. Deselect and Invert never do: Deselect also clears lines and symbols picked with the Arrow, and inverting nothing is how you select everything.
A selection belongs to its document. Switch tabs and it stays behind; switch back and it is where you left it. A new document starts with nothing selected.
Transforming a rigged drawing moves what you see. With the character standing in a pose, Ctrl+T shows the posed drawing under the handles, a box drawn over a posed line catches that line, and confirming moves the posed picture by exactly what you did to it β a straight line stays straight across a joint. The drawing stays rigged afterwards and keeps following the pose. The skeleton itself does not move with the transform; to move a whole character, grab a bone in pose mode instead. Three small limits: a transformed line loses any curve handles it had (the points are still exact, and a later fit brings the handles back); the band scale adds no extra points on its dividers to a posed drawing; and a point sitting exactly in a joint folded almost flat stays where it is, because there is no rest position that would put it where you dragged it without throwing it off in every other pose. A line made of only a few points that spans a joint gains points when it is transformed, so that it keeps its shape between them; drawn strokes already have plenty.
Copying lines out, and pasting them back
Ctrl+C copies what you have selected; Ctrl+V pastes it onto a new layer (unless a layer row has focus β then they are the layer keys above). It works with whichever tool made the selection β a box or lasso from the Select tool, or lines picked with either arrow β and whatever the drawing is made of: lines, and pixels too, which is what a merged, imported or flattened layer holds. The two gestures mean slightly different things, on purpose:
- A region copies what you boxed. Draw a box across half a line and the paste shows that half; draw it over pixels and the paste is the pixels inside the box, cut at its edge. The line is not cut in two to do it: it travels whole with your selection carried along as its clip, so the record still holds one line, and a later change of mind can still reach the rest of it.
- Picked lines copy whole. Clicking a line with the arrow means that line, all of it, however much of it your last box happened to cover.
When both a box and some picked lines are up, the box wins β the same order the transform tool uses, so the two never disagree about what "selected" means.
The paste lands on a new layer directly above the one you are on β inside its folder if it is in one β in the place it was copied from. In place because the usual reason to copy a drawing is to carry it to another frame or another shot, where landing anywhere else means lining it up by hand every time; move it afterwards with any of the tools if that is what you wanted. On its own layer because a paste is not part of the drawing underneath it until you say so β merge it down when it is.
The clipboard reaches between open documents. Copy in one tab, switch to another and paste. It is Lightbox's own clipboard, not the system one, so nothing leaves the app and nothing arrives from outside it.
Ctrl+X cuts: it copies, then takes the same content out. A boxed region is cleared β lines and pixels alike β and lines crossing the edge keep the part that was outside; picked lines are removed entirely.
A selection with nothing in it copies nothing. The status line says so, and neither key reaches past the selection to the rest of the drawing.
Ink you have erased is never copied. A line rubbed out along its whole length is not on the canvas, so a box drawn over where it used to be picks up nothing β the same rule that stops it being clicked, or moved.
With nothing selected, Ctrl+C copies the whole layer β every drawing and hold, as Copy layer does, or the folder or layers picked in the docker β and Ctrl+X copies it and empties it: the layer stays where it is with its mask, effects and settings, and its drawings go, as one undo step. Ctrl+V always pastes onto a new layer above the active one, taking whichever you copied most recently β lines, a layer, or cels.
Cels are the exception, by name. The timeline's right-click Copy and Paste are always the cel, and paste onto the cel you opened the menu on. With nothing selected on the canvas, a cel with keyboard focus, or several cels picked in the timeline with the playhead's among them, make Ctrl+C and Ctrl+X mean those cels too; Ctrl+V still pastes them onto a new layer, starting at the playhead. The verbs also have menu entries β the lines on the Select menu, the cel on Animation β and Configure β Shortcuts has keys for the line-only forms if you would rather they never meant the layer.
To crop the paper to a selection, see Cropping under Documents and projects β it is on the Image menu, with the two resizes, because it changes the page rather than the selection.
Hold Ctrl inside a selection and drag to move what is in it, without reaching for a tool. The pointer becomes the four-way arrow as soon as you cross into the marching ants, so you can see that you have got it before you press β and it keeps that shape for the whole drag. Shift holds the move to one axis, guides catch it, and the whole thing is one undo step. Outside the ants, Ctrl is still the eyedropper it always was; the boundary is the selection edge and nothing else. It moves the lines the selection holds rather than cutting the pixels out, so a stroke crossing the edge travels whole. Ink you have erased stays erased: eraser marks caught by the selection travel with the lines they carve, but what they rubbed out of the lines that stay behind does not come back β moving a selection never resurrects anything, in the drag or on apply. The selection takes what you can see: a line you rubbed out entirely is not on the canvas, so a box drawn where it used to be does not pick it up β and a half-erased line goes by where its surviving ink sits, so boxing the visible end moves it even when the rubbed-out part lay elsewhere.
Ctrl+T starts a transform. The gizmo gives move, scale, rotate and a draggable pivot; Perspective mode gives four free corners. The whole session is also under Edit βΈ Transform β start it there, and apply, cancel, mirror either way, switch perspective on or reset without leaving the menu; the same verbs are on a right-click over the gizmo.
Band mode: stretch part of a drawing, keep the whole
The legs are too short and the figure is the right height. A scale is the wrong tool β it changes the whole sketch β so Ctrl+Shift+T switches the gizmo to band mode, where the box is pinned and you redistribute the space inside it instead. It is also on the gizmo's right-click menu, and it is rebindable like any other shortcut.
| To do this | Do that |
|---|---|
| Place a line across the box | Click where it goes β at the hip, say |
| Place an upright line instead | Shift+click |
| Move a line | Drag it |
| Take a line away | Alt+click it |
| Apply | Enter, as ever |
Drag the hip line up and the legs lengthen while the torso shortens: the outer edges of the box never move, so the figure keeps the height it had and only its proportions change. That is the whole idea, and it is why the corner and edge handles do nothing in this mode β a box that resized would be a scale again.
Only the two bands either side of the line you dragged can change. Place two lines and you have three bands; move the lower one and the top band is untouched, because it is held by a line that did not move. Nothing you did not touch moves.
Placing a line changes nothing on its own, so you can lay out the bands you want before touching any of them β and pressing Enter without having dragged anything leaves the drawing exactly as it was, with no undo step to step back over.
Two things worth knowing:
- A line weight is not stretched. A scale carries brush size with it; a band does not, because it has a different factor per band and only along one axis. A lengthened leg keeps the weight it was drawn with, which is what correcting a proportion means.
- Imported pixels stay where they are. A band scale cannot be written as a single transform, so there is nothing for a raster layer's pixels to be resampled through. Strokes move exactly; imported pixels do not, and the status line says so when a drawing in scope has any.
The band layout is a gesture rather than part of the document: apply, and the lines are gone. If you want a proportion you can come back and adjust next week, that is what a rig is for.
Cage mode: bend part of a drawing
The arm is straight and should curve; the jaw wants pushing out a little. Neither is a scale and neither is a band. Ctrl+Alt+T switches the gizmo to cage mode: a lattice of handles covers the box, and dragging one bends the drawing smoothly around it while the rest stays put. It is on the gizmo's right-click menu and in Tool options beside Bands, and it is rebindable.
| To do this | Do that |
|---|---|
| Bend the drawing | Drag a handle |
| Bend a smaller part | Raise Grid in Tool options (2 to 6 cells a side; the lattice is laid out afresh) |
| Start over | Reset transform on the right-click menu |
| Apply | Enter, as ever |
A handle moves what is near it and nothing two cells away. The box's own edges are pinned, as in band mode, and nothing outside the box moves at all β so a cage over a figure's arm leaves the torso alone. The handles you have not touched hold their part of the drawing where it was.
Lines made of few points gain the points they need to bend, so a straight stroke across the cage comes out as a curve rather than a kink. Line weight is not changed. Imported pixels follow the cage: a photo or a scan under the lines is resampled through the same mesh you watched while dragging, which is the one way a band scale and a cage differ β a band leaves pixels where they were, a cage takes them along.
On a rigged drawing the cage bends the posed picture you are looking at, like every other transform now, and the drawing keeps following the rig afterwards.
Like the bands, the cage is a gesture: apply, and the handles are gone. Pressing Enter with every handle still at home changes nothing and leaves no undo step.
Scaling holds the side you did not touch. Drag a corner and the opposite corner stays put; drag an edge and the opposite edge does β which is what every other tool does, and what your hand expects when it pulls one side of a box. Hold Alt to scale from the pivot instead, growing all sides at once, and you can press or release Alt in the middle of the drag. Shift keeps the proportions, as ever, and combines with Alt. The drawing moves with the gizmo β you see the result while you drag, not after you commit. The marching ants step aside while the gizmo is up, and come back around the drawing in its new position when you apply β the gizmo boxes the ink it is moving, the ants trace the region you selected, and those two are rarely the same shape, so showing both at once reads as two tools arguing rather than one tool working. Your selection is not lost while it is out of sight: it moves with the drawing, so a second transform straight afterwards lines up with the first. Cancelling brings it back where it was.
Dragging with the Move tool keeps its ants, because there is no gizmo there and the outline is the only thing showing you where the pixels are going. The session's controls β scope, sampling, perspective, mirror, Reset, Apply and Cancel β live on their own page of the Tool options docker, which opens by itself when the session starts, so they are never off screen while a transform is live. Enter applies, Esc cancels, from the keyboard as always.
While a transform is up, the canvas belongs to it β every press on the drawing goes to the handles, whatever the toolbar says. Picking a tool ends the session and discards the drag, on the grounds that reaching for the brush means you are done transforming. Nothing is written to the drawing that way, so an accidental tool press costs you the drag and never the artwork. Holding Ctrl for the eyedropper is a borrow rather than a choice and leaves the transform alone.
Ctrl+Z steps back through the transform, not through the drawing. Each drag of a handle, each mirror, each Reset and each switch between box, perspective, bands and cage is one step: Ctrl+Z takes the last one back and Ctrl+Y puts it back again. With every step taken back the box is as it opened, and Ctrl+Z goes no further β nothing you drew before the transform can be undone from inside it. Esc is the way out. The Edit menu says Undo transform step while this is so.
Applying makes the whole transform one undo step, however many tweaks it took, so Ctrl+Z straight afterwards puts the drawing back exactly as it was before you pressed Ctrl+T.
Moving to another frame, layer or document applies the transform, as Enter would, and ends the session. It is applied to the drawing it was on, not the one you moved to, and it leaves the same single undo step. A box you never moved applies nothing and leaves no step. Picking a row in the History docker cancels the transform instead β a row is a different state of the document, and the box was drawn round this one.
Scope decides what moves:
| Scope | What moves |
|---|---|
| This drawing | The drawing at the playhead on this layer (the default) |
| This frame, every layer | The drawing at the playhead on every visible layer |
| This layer, every frame | Every drawing on this layer, whatever frame it is on |
| Marked cels | The cels marked on the timeline, holes and all |
| Everything | Every drawing on every layer |
A drawing held across several frames is moved once, not once per frame. Whenever the scope reaches past the drawing in front of you, the status line says how many drawings will change β Transforming 12 drawings on Walk β since most of them are on frames you cannot see while you drag.
Every frame of a layer at once
The character is too small, on every frame of the cycle. Press Ctrl+Shift+Alt+T (or Edit βΈ Transform βΈ Every frame of this layer) and one box goes round every drawing on the layer. Scale, move or rotate it once and every drawing follows; Enter applies it as one undo step.
There is one pivot for all of them, and that is the point: drag the pivot to the feet, scale, and the figure grows and so does its motion β a jump keeps its height in proportion to the jumper and the feet stay on the ground line. Scaling each drawing round its own middle would grow the figure while the jump stayed the size it was. Line weight scales with the drawing, as with any transform.
Turn on onion skin to see the rest of the cycle while you drag. The ghosts of the drawings in the box move with it, so the frames either side show the size they are about to become, not the size they were. That holds for any transform that moves whole drawings across frames; with a selection up, only part of each drawing moves, so the ghosts stay where they are rather than show lines moving that will not.
This is a gesture, like the Move tool's Ctrl-drag: the Scope control reads This layer, every frame while it is open, and the next Ctrl+T is back on whatever scope you had set. Pick another scope while it is open and that one takes over. With lines picked it does nothing and says so β a picked line lives on one drawing, so deselect first (Ctrl+D).
The Move tool does not read or change the scope either β a plain drag moves this drawing and a Ctrl-drag every drawing on the layer, for that drag only.
Ramp it over the frames
The cycle walks on the spot and should cross the screen. Start a transform over every frame of the layer (or mark a run of cels and pick Marked cels), tick Ramp on the Transform page, and drag the box to where the last drawing should end up. The first drawing stays where it is, the last gets the whole box, and every drawing between gets its share β a quarter of the way along at a quarter of the way through. The same works for a scale (a push-in that grows over the shot) or a rotation.
| To do this | Do that |
|---|---|
| Make the move gather pace, or settle | Pick Ease in, Ease out or Ease in and out beside Ramp; Even is the same step every drawing |
| See the whole ramp before applying | Turn on onion skin: each ghost shows its own share |
| Apply | Enter β one undo step for every drawing |
What to know:
- The box is the last drawing. The drawing under the playhead shows its own share, so on the first frame it does not move at all while you drag β watch the ghosts, and the status line says what share this frame gets.
- Holds keep their timing. A drawing held for two frames takes the share of the frame it starts on, so a cycle on 2s still moves on 2s.
- A drawing shown in more than one place is split. A loop that shows its first drawing again at the end cannot have it at both ends of a walk, so each later place gets its own copy, and the status line says how many were made. Drawings outside the ramp are never changed β one that also shows outside it is copied at every place inside it, and when the last drawing is held on past the marked cels, the first frame after them keeps it as it was. Ctrl+Z puts it all back.
- Box only. Perspective, bands and the cage apply in full, and a ramp needs whole drawings, so it is not offered while a selection is up. A mirror cannot be ramped β it would pass through a flat line on the way β and Enter says so rather than mirroring everything.
- The nearest ghost each side follows the drag live; the rest catch up when you pause. Stop moving for a moment and every ghost stands at its share, so you read the whole ramp each time you stop β and however deep the onion skin, the drag itself only redraws the two nearest.
- The ramp switches itself off when the transform ends; the next one starts without it.
Pick several layers in the Layers docker first and Ctrl+T moves them together β the box goes round the drawing at this frame on each picked layer, and one drag, scale or rotation moves them all, as in Krita. It is this frame only: the same drawing on each layer, not the frames before or after it. A picked layer that is hidden or locked stays where it is, and the status line says how many did. A picked layer that is holding an earlier drawing at this frame gets a drawing of its own when you apply, exactly as the layer you are on does, so the frames that borrowed it are left alone. The Move tool follows the same picks.
With a selection up, only what is inside it moves. Draw a box across half a line, drag, and half the line goes β the half you boxed. The other half stays exactly where it was. It works with any of the selection shapes, and however little of a drawing you take: a marquee over one corner of one line moves that corner.
Pixels move too. A layer that is pixels rather than strokes β an imported image, a PSD layer, or a layer a merge had to flatten β moves the same way: the pixels inside the selection go, the rest stay, cut at the selection's edge. Ctrl+T, the Move tool and every transform mode (perspective, bands, the cage) treat them alike, and a layer holding both pixels and strokes moves both.
What counts as inside is the ink, not where the pen went. A selection made on one layer works on every layer you then pick, and it takes whatever paint of theirs lies inside it: select the colour layer's pixels, switch to the line layer, and Ctrl+T takes the part of each line that overlaps the colour; draw a box inside a filled area and the colour in the box moves. The selection itself belongs to no layer. The magic wand selects exactly what it matched β with sample all layers on, a wand clicked on the colour stops at the lines and the shading drawn over it, so they are outside the selection and stay put. To move them with the colour, draw a lasso or marquee round the whole thing, or let the wand sample only the colour layer.
The line is not cut in two to do it, which is the same answer copying a region gives and for the same reason. The record still holds one line, carrying your selection as its clip, so nothing about the mark itself changes β a brush with scatter or jitter lands its grain exactly where it always did, and a later change of mind can still reach the rest of the line. A stroke that sits wholly inside the selection simply moves, with no clip and nothing left behind.
A gradient is treated like anything else. Move a selection on a layer with a gradient on it and the ramp inside the marquee travels too β so a marquee over a gradient sky leaves a rectangle of sky that has shifted with whatever you moved. If that is not what you want, put the gradient on its own layer and transform the layer above it.
An eraser travels with the ink it carved. If you rub a bite out of a line and then move the line, the bite goes too β and if you move only part of it, the part left behind keeps its own edge. Rubbed-out paint never comes back, whatever you select and wherever you drag it.
Lines picked with the Arrow can be transformed too, and then the scope is this drawing, because a line lives on one. If you have both an area selected and lines picked, the area wins; the transform says which it took when it starts, so you are not left guessing.
The handles fit the drawing you can see. A drawing you have reworked still remembers what you rubbed out, and the box ignores it β so the handles sit round the art rather than round its history. An imported drawing transforms too, even though it is pixels rather than strokes; with a selection up the imported pixels stay put, so the box shrinks to the lines that are actually going to move.
Because strokes are geometry, a transform is lossless: rotating and rotating back leaves no softening.
A line painted under a selection keeps its clip, and the clip comes with it. Painting inside a marquee leaves the line holding that shape as a stencil for ever after, which is what makes a reload paint what you saw. Move or rotate that line and the stencil turns with it, so the mark you end up with is the one the drag was showing you β rather than the old stencil cutting the new position, which used to leave that one line looking as though it had jumped.
Guides
View β Guides places rulers, grids, isometric axes, vanishing points and the construction aids β a character height scale, an eye-line, a horizon. None exist until you place one, and a document that never uses them carries no guide machinery at all.
| Guide | What it constrains |
|---|---|
| Horizontal / vertical ruler | Strokes drawn along it come out straight |
| Grid | Points snap to its intersections β corners, shapes, the starts of strokes |
| Isometric | Three axes at Β±30Β° and vertical, from one origin you can move once |
| Vanishing point | Strokes radiate from it. One is one-point perspective, two is two-point, three is three |
| Character height scale | Points snap to its head-unit divisions β chin on the fifth head, knees on the second |
| Eye-line / horizon | A horizontal ruler that wears its name, so a shared rig reads at a glance |
They do two different things, and knowing which is which is the whole trick:
- A grid snaps points. Each point independently goes to the nearest intersection. That is what you want when you are placing things.
- A ruler or vanishing point constrains a stroke. The first part of your drag says which direction you meant; once you have travelled far enough to be believed, it locks to the guide that matches and holds the rest of the stroke on that line. That is what you want when you are drawing along something.
A ruler only takes a stroke that starts on it. Begin within the snap tolerance of a guide line and head along it, and the stroke is ruled onto that line. Begin anywhere else and the stroke is yours, however parallel to the ruler it runs β a ruler is a line in a place, not a direction the whole canvas obeys. A grid never rules a stroke at all: it puts the start on an intersection and lets go. Vanishing points and isometric rigs are the other kind β they have no single line to be near, so they constrain a stroke wherever it is drawn.
Locking once is deliberate. Re-deciding every moment would mean a slightly wobbly hand flicking between two vanishing points mid-stroke and the line kinking. Draw across every guide and none of them takes it β a guide that grabs strokes you meant freehand is a guide you turn off.
The ruler decides the direction, not how far: the stroke still goes where your hand went. β in the shortcut bar turns snapping off without removing anything, and hiding a guide does not stop it snapping β those are two switches, because hiding a rig to look at the drawing under it is something you do constantly. The snap switch is remembered: turn it off and it is still off the next time you open Lightbox, in every document, until you turn it back on.
What snaps: anything you aim, and nothing you trace. A point you place deliberately goes to the guides β a stroke's start, a shape's corners, a gradient's axis, a pen node, a polygon selection's corners, a rectangular or elliptical marquee, and a point dragged with the white arrow. A line you draw by hand does not, past its first point: a lasso and the middle of a brush stroke follow your hand, because a guide pulling at every sample would fight you the whole way round.
Moving something snaps its edges and its middle, rather than wherever you took hold of it β so lining a drawing up against a ruler means dragging it near the ruler, not grabbing its exact edge first. The corners find rulers and grids; the centre finds a vanishing point. Whichever is nearest is the one that lands, and if nothing is within reach the drawing goes exactly where your hand put it. This is the Move tool and dragging lines picked with the Arrow; the transform handles do not snap yet.
Previews tell the truth about this. The pen's rubber band runs to the place the node will actually land rather than to the pointer, so what you see before the click is what the click does.
Holding Shift beats a guide where the two disagree. Shift asks for an angle β a level gradient, a square marquee β and a guide asks for a place; if a guide pulled the end back off the angle you had just asked for, one of the two would be lying. So with Shift down the angle wins, and the guides keep the other end.
Guides are saved with the document, like the camera, and drawn over the artwork, translucent β under it they would vanish the moment they crossed an opaque background layer. The snapped points are what the stroke records, so moving a guide afterwards never moves a line you have already drawn.
Adjusting a guide
Pick up the Move tool or the Arrow and the rig wakes up. Every guide on the canvas lights faintly, the one under the pointer brighter, and the ones you click brightest of all β and a grid or an isometric rig, which is grabbed at its anchor rather than on its lines, shows that anchor as a small square you can aim at. With any other tool in hand the guides go back to being scenery you draw over, so nothing glows at you while you are drawing.
Two tools reach guides, and they do the same things to them. The Move tool is the one you reach for when the job is the rig; the Arrow picks guides alongside the lines, symbols and anchors it already picks. Whichever you use, a guide behaves the same way β the same click, the same drag, the same options, and a guide you locked is out of reach of both.
Clicking a guide points the tool options at it, in the quick bar and in the Tool options docker, because moving a guide and changing one are the same intention reached for with the same hand. Clicking the drawing lets go again.
Shift adds a guide to the selection and Alt takes one out, the same as everywhere else on the canvas, and dragging any guide in the group moves all of them together as one undo step. With more than one picked the options go quiet: they set a guide's numbers β this grid's pitch, this chart's head count β and there is no single guide for them to mean. The group still moves.
The selection is handed between those two tools and dropped by every other one. Switching between the Move tool and the Arrow keeps what you picked, because both can act on it; reaching for the brush lets it go, so nothing stays lit over a drawing you are painting.
| The selected guide | What you can set |
|---|---|
| Any guide | X and Y, in document pixels β the typed half of the drag, for a horizon at exactly y=540 or two vanishing points exactly as far outside the frame as each other |
| Guide line, grid, isometric | Angle, in degrees clockwise from horizontal |
| Grid | Cell size β the pitch of this lattice |
| Character height scale | Head height and Heads β the same two numbers the top-rung drag changes, typed exactly |
| Vanishing point | Rays β how many lines are drawn out of the point |
| Any guide | Shown, Snaps and Locked β and Remove, or select it and press Delete |
Every one of those changes this guide, on this drawing, and every one is an undo step. A locked guide ignores the fields exactly as it ignores a drag.
The ray count is what you see, not what you can snap to. A vanishing point constrains every direction through it whatever the fan is drawn at β fewer rays to see the drawing through it, more to read the perspective, and the strokes land in the same place either way.
"Set as default" is the other half. It is the deliberate act that also changes what the next guide of that kind is made from: a grid's cell size, a vanishing point's fan, a height scale's proportions. Without it nothing you do here reaches beyond the drawing in front of you, which is the point β a preference that rewrote itself every time you nudged one guide would not be a default. The same values live in Edit β Configure β Guides and grid if you would rather type them there.
A height scale saves a proportion rather than a head height in pixels β "six heads standing in seven tenths of the canvas" β so the same default still lands as a figure on a scene of a different size. Resizing a document afterwards does not move a chart already on it; its size is document data by then.
The character height scale, and the named lines
View β Guides β Add character height scale stands a head-unit chart on the canvas: a post with a rung per head and a count beside the top β "6 heads". It is how animators keep a character on model, and it behaves like one thing because it is one:
- Drag the top rung to resize the character. The divisions follow β a head count is a proportion, and a taller character is not more heads. The whole pull is one undo step. (Grabbing the post anywhere else moves the chart.)
- Its divisions snap, all the way across the canvas: put the chin on the fifth head wherever the character is standing. Above the top and below the ground there is nothing to snap to β the chart's extent is its point.
- Edit β Configure β Guides and grid lists every height scale on the document, where one head's height and the head count can be typed exactly.
Add eye-line and Add horizon place ordinary horizontal rulers that carry their names on the canvas. They constrain exactly what a ruler does; the label is the feature β on a rig somebody else set up, or pulled from a guide set, a line that says "Horizon" needs no archaeology. Any guide that has a name wears it the same way, and a guide set keeps the names of the guides in it.
Rulers, and placing a guide by eye
Edit β Show rulers (Ctrl+R) puts a ruler along the top and left of the
canvas. They count in document pixels, they mark every guide that crosses
them, and a line slides along both as you move the pointer β knowing where you
are without stopping to read a tick is most of what a ruler is for.
Drag out of a ruler to place a guide. Out of the top one for a horizontal guide, out of the left one for a vertical one; the guide follows the pointer while you aim it. Let go back over the ruler and it never existed, which is both how you delete one and how you get out of a drag you did not mean.
A guide on the canvas is picked up with the Move tool or the Arrow, wherever the rulers are. The cursor changes when you are on one and the guide itself lights up; there is nothing floating over the drawing to click instead. The whole drag is one undo step, not one per twitch of the hand.
Shift holds a dragged guide to one axis β whichever the drag has gone furthest along, measured from where you picked it up, so a horizon slides along without drifting off its height. It is the same promise Shift makes on the brush and on the Move tool's content drag, and it covers the group drags too: guides, reference boxes, anchors and hitbox shapes moved together all hold the same way. Press or release Shift in the middle of the drag and the thing follows β the lock is read from the keys as you move, not once at the press.
The tool is the switch for all of this, on purpose: grabbing a guide and drawing along one are the same gesture in the same place, so something has to say which you meant. With any other tool in hand a guide is scenery you draw over and nothing can nudge it by accident. Hiding or locking the guides overrides both of them, because both mean "leave the rig alone".
| Edit menu | Key | What it does |
|---|---|---|
| Show rulers | Ctrl+R |
The strips, and with them the drag-out |
| Show guides | Ctrl+; |
Take the rig off the screen. It still snaps |
| Lock guides | Ctrl+Alt+; |
Pin them where they are, rulers or no rulers |
β and π in the shortcut bar are the last two, and they appear there only while the rulers do β off the rulers, neither could change anything.
Rulers, guide visibility and the lock belong to the workspace, not the document: they are how your screen is arranged, so they save, reset and switch with everything else, and opening somebody else's file never rearranges them. They also survive a restart β rulers left up are up again next time, saved workspace or not.
Several at once: Shift-click to add guides to the selection, Alt-click to
take one out, and dragging any of them moves the whole group together as one
undo step. Either guide tool does both halves β picking and dragging are the
same tool's job, not one tool's each. Ctrl+A with the Arrow in hand takes
every guide on the document at once. The lock still holds: locked guides stay
put whether they were selected or not, which is the point of locking them.
Guide sets β the same rig on every drawing that needs it
A character's height lines, a street's perspective rig, an isometric grid the whole level shares: guides you will want again on the next drawing. A guide set is a named copy of a document's guides, kept in the project β which is why this needs one; a loose document has nowhere to keep it.
View β Guides β Guide setsβ¦ is where sets are made and managed. Place your guides on the canvas first β the canvas is the guide editor, and a set is work you already did, the way a template is β then name them and save. The same window renames a set, refreshes it from the open document, or deletes it (deleting also takes back anywhere it was shared).
View β Guides β Add from set pulls a set into the open drawing. The guides arrive as ordinary document guides β one undo step, and yours from then on: dragging one afterwards changes this drawing only, never the set, and never another drawing that pulled from it. Refresh the set deliberately from the editor when the rig itself should change.
A set lands on the paper it arrives at. Sets remember the document size they were saved from, so a rig drawn on a 4K sheet and pulled into a 1080p one arrives at the same place in the frame at the same share of the height β a six-head knight that filled 70% of the tall sheet fills 70% of the small one, still six heads, still standing on the same ground. That is what keeps a character the same height across a project whose documents are not all one size.
Where the two documents have different shapes rather than different sizes, height wins: everything is scaled by one factor taken from the height, and only the anchors move sideways to keep their place across the frame. A line keeps its angle, a grid stays square, and an isometric rig stays isometric β which scaling the two axes separately would quietly have ruined.
Sets saved before this existed have no size recorded and arrive exactly as they always did, in the pixels they were saved at. Save one again from an open document to teach it its paper.
Sharing decides what "Add from set" offers. Out of the box every set in the project is offered to every document. In the project window, right-click a folder to share a set onto it β from the first share onwards the project is scoped, and a document is offered what its own folder (or the folders above it) declares: the knight's height guide stops appearing in the goblin's menu. This is the same scoping palettes, gradients and brush tips use.
A shared set is also what a new drawing opens with. Make a document while you are in the knight's folder and the knight's guides are already on it, fitted to whatever size you chose β which is the point of declaring a height chart on a character at all, since Add from set on every new drawing is the step that gets skipped. The nearest share wins, exactly as a palette resolves: a drawing under the knight gets the knight's rig rather than a project-wide one, and only that one. Nothing is applied in a project where you have never shared a set, so this follows a deliberate share and never a default.
The guides a drawing opens with are part of it rather than an edit to it: the drawing does not read as unsaved work you did not do, and undo before your first stroke will not take them away. Delete them like any other guide if a particular drawing does not want them.
Guide defaults, and the guides already placed
Edit β Configure β Guides and grid holds what new guides are made from β a grid's cell size, a vanishing point's ray count, a height scale's head count and its share of the canvas height β and how close a point has to come to a guide to be pulled onto it. It also lists the grids and height scales already on the document, where a grid's pitch and angle, a scale's head height and count, and each one's drawing and snapping can be changed after the fact β each one an undoable step.
Changing the default cell size never touches a grid that already exists. Once a grid is placed its spacing belongs to the document, and a preference must not reach back into work you have already done against it.