Skip to Content
Starlight

The node graph

The node graph shows the scene as a diagram: what feeds each surface, in what order, and on which clock. Press g, or pick Window › Graph, to show or hide it.

Read the graph

Each box in the graph is a node: one layer, one transform, one effect, one comp or one surface. Wires show where each result goes. Content flows from left to right.

  • Left: the sources. These are the layers that draw something: text, image, video, shape, gradient.
  • Middle: transforms and effects. A transform places the layers inside it: position, scale, rotation. Each effect shows as a small chip in a chain after its node.
  • Right: the surfaces. A surface is one LED output, such as N or STAGE. It has a panel icon and its size, for example N · 6400×1600.
  • Upstream means toward the sources. Downstream means toward the surface.
  • Comps: a comp is a named group of layers that you can place on surfaces. Its node has a stacked-layers icon and a ▸N badge. N is the number of layers inside (Glossary).
  • Used by: a surface can use a comp inside its own layers, for example its slice of the panorama. At the top level, a faint dashed link goes from the comp to each place that uses it. This link is not a feed. Hover the comp to read used by N, E1, …. A comp with no links and no feed is not used.
  • Layer order: nodes stack from top to bottom, and the top node is in front. Where two or more layers meet, each wire has a tag with its number (1 = front) and its blend, for example 1 overlay. Hover the tag to read more.
  • Clocks are the round dials in a lane on the right (Use the clock dials).
  • When the top level has only comps and surfaces, two faint labels name the columns: comps and surfaces.
  • Hover an icon to read what it means. For example, the surface icon says a project output: what flows in renders here.
  • When you zoom far out, small text (such as a surface's size) and the dot grid hide.
  • A long name is cut short in the middle with …. Its number at the end and its file extension stay, so different versions of one file stay easy to tell apart, for example image pix…_00012.png. Hover the node to read the whole name. The layers view, a comp's layer list and the Assets panel shorten names the same way.
  • To move the graph to another edge of the window, see Move the graph.

Small marks on nodes and wires.

MarkMeans
Four arrows on a transformThis transform only places the one layer it holds. It has that layer's name
◑The layer uses a blend other than normal. Hover the node to see the blend
Hollow chip with a dotted edge, and ◧An adjustment layer: it changes the layers below it
Two inputs: content and matteA matte node: the shape in its matte input cuts its content, like a mask
Small ring at a node's right endThe layer has its own animation. Hover the ring to see what times it. An amber ring: its clock does not fit the scene loop a whole number of times
Dim, hollow outlineThe node is bypassed (switched off). A dim use or clone: its comp is bypassed
Small round switch on the left edgeBypass. Faint while the node is on, bright while it is off. Click it to switch (Turn a node off)
A card behind the node, and ▸NA comp you can open. Hover ▸N: it says N layers inside · double-click to open
Output mark on a comp's rootInside a comp: the place where the comp's result leaves
Dot on a comp's right edge or a surface's left edgeA feed port (top level only). It is filled when a wire is plugged in (Make a comp)
Small arrowhead where a wire meets a surfaceContent flows into the surface here
Faint dashed wireA link: a clock to the nodes it times, or a comp to each place that uses it
Dashed wire with an arrow, between dialsA trigger: the clock at the arrow starts at a cue of the other (Triggers)

Select and navigate

Select nodes to edit them, and move around a large graph.

ToDo this
Select a nodeClick it
Clear the selectionClick empty graph
Add a node to the selection, or take it outHold ⇧ / Shift, ⌘ / Ctrl or ⌥ / Alt and click it
Select with a boxRight-drag across empty graph. Hold ⇧ / Shift to add to the selection
PanDrag empty graph, drag with the middle mouse button (also over a node), scroll with ⇧ / Shift held, or swipe sideways
ZoomScroll or pinch. The zoom centres on the pointer
See the whole graphPress ⇧1 / Shift+1 with the pointer over the graph, double-click empty graph, or click the fit button (bottom right, tooltip fit graph)
Walk from node to node→ downstream, ← upstream, ↑ ↓ to the node above or below, also in another branch
Move a nodeDrag it
Delete the selected nodes⌫ / Backspace or Delete
Go back to the automatic layoutRight-click empty graph › reset layout
  • On a Mac, Control-click also adds or removes a node, and it never opens the menu. A right-click or a two-finger click opens the menu.
  • With a modifier key held, a click on empty graph keeps the selection.
  • Surfaces, comps, clock dials, wires and notes are selected one at a time.
  • The box selects layers and effect chips only. It skips surfaces, comps, clock dials and notes. A box over nothing (without ⇧ / Shift) clears the selection.
  • Right-drag selects and left-drag pans. Many other apps do the opposite. A right-click that moves less than 6 px opens the menu.
  • A middle-button drag pans the graph, also over a node or a note. It never selects or moves a node.
  • If you drag or right-click a node in the selection, the action applies to every selected node. This works for move, duplicate, delete, bypass and colour. It is one undo step.
  • The arrow keys follow the pointer. With the pointer over the canvas, they nudge the selection (Nudge with the arrow keys). Anywhere else, they walk the graph. With nothing selected, or over a dial, ← and → step one frame (10 with ⇧ / Shift).
  • ↑ and ↓ select the nearest node above or below. They also go to another branch: another chain of nodes, like the next row in a list of files.
  • A node in the same column comes first. For example, two layers each have a transform and a video. ↓ on the top video selects the other video, in the next branch.
  • If no node is above or below in that column, ↑ and ↓ select the nearest node in any column. The nearest row comes first, then the nearest node in that row.
  • At the top or the bottom of the graph, ↑ and ↓ stop. They do not go round to the other end.
  • The arrows use the graph as you see it. If you moved a node, they use its new place. They never go into a comp that you see as one node. To walk the layers of a comp, double-click it to go inside.
  • When you select a node that is off screen, the graph scrolls to show it. A new node is shown in the middle.
  • The graph zooms from 15% to 300%.
  • A node's place is saved with the scene. Moving a node is an unsaved edit, and ⌘Z / Ctrl+Z undoes it. Pan and zoom are saved for each scene, but they are not part of undo.
  • Every key and mouse action for the graph: Keyboard & mouse › Node graph.

Add a node

Add a new layer, such as text, a shape, a gradient or an image. In the graph, each layer is a node.

  1. Right-click empty graph, near the surface or comp where you want the layer.
  2. Point to create and pick a kind, for example text or gradient.
  3. The new layer is selected and opens in Properties. Change it there.

The create menu lists these kinds: text, rectangle, ellipse, line, pen, gradient, image, cutout, audio, video, clone, adjustment layer and clock. Kinds that the project's extensions add come after them. For text and clock, the command palette opens first: type the words or the clock's name, then press ⏎ / Enter. See Layer types.

Other ways to add a node.

  • Next to a layer: right-click the layer, point to add sibling and pick a kind. The new layer goes on top, on the same surface or in the same comp.
  • The command palette: press / or ⌘/ / Ctrl+/. Type a kind (text, rectangle, gradient) and press ⏎ / Enter. For text, type the words and press ⏎ / Enter again.
  • From a file: drop it on the graph (Drop a file or asset).

Where the new layer goes.

  • create uses the comp or surface nearest to your click. On a surface, the layer goes into the surface's own layers.
  • Inside a comp (Step into a comp), the palette adds the layer to that comp.
  • At the top level, the palette adds it next to the selected node. With nothing selected, it uses the surface the pointer was last over on the canvas. If there is none, it uses the first surface.
  • When the palette picks a surface in this way and that surface shows a comp, the layer goes inside the comp.
  • The layer appears where the pointer last was on the canvas, if that point is on its surface. If not, it appears in the middle.
  • A new text, shape or gradient starts in the brand kit's primary colour.
  • A new image, video or audio layer uses the project's first file of that kind. A new clone repeats the first image. To start from another file, drop that file on the graph instead.
  • clock adds a named clock dial, not a layer. Its length is the scene's loop. pen lets you draw a path on the canvas (Draw a path or a mask).
  • Each add is one undo step.

Drop a file or asset

Make a layer from an image, video or sound file.

  1. Drag a file from Finder (Mac) or Explorer (Windows), or drag a tile from the Assets panel.
  2. Drop it anywhere on the graph.
  3. The new layer is selected and opens in Properties.
  • The layer goes where the palette would put it (Add a node). That is inside the comp you are in, or next to the selected node.
  • A sound becomes an audio dial. An image or a video becomes a new layer.
  • A file from Finder or Explorer is imported into the project first (it is copied into Assets at once). The new layer is an unsaved edit, like any other: press ⌘S / Ctrl+S to save it, or ⌘Z / Ctrl+Z to take it away. A tile from Assets works the same way.
  • To put media on one particular surface, drop it on that surface on the canvas instead (Drop media on a surface).
  • Files you can drop: PNG, SVG, JPG, WebP, GIF, MP4, WebM, M4V, MOV, MP3, WAV, M4A, AAC, FLAC and OGG.
  • Drop one file at a time. If you drop several, only the first one becomes a layer.

Make a comp and place it on surfaces

Group layers into a comp, then show that comp on one surface or on many surfaces.

  1. Select the layers that you want in the comp. They must all be in the same list: the same surface, comp or transform.
  2. Right-click one of them and pick group into subtree.
  3. The layers move into a new comp named subtree (then subtree-2, and so on). One comp node takes their place, and Starlight selects it.
  • The layers keep their own places inside the comp, so the picture does not change.
  • The layers do not need to be next to each other. The comp goes where the backmost of them was. If other layers were between them, the layer order changes.
  • If the layers are not in the same list, nothing happens. You cannot group the two inputs of one matte node.

Show a comp on a whole surface.

  1. Go to the top level of the graph. To get there, click the scene name in the breadcrumb (the row of level names at the top left).
  2. Find the comp's node. Its output dot is on its right edge. Its tooltip says output · drag onto a surface to feed it.
  3. Drag from the output dot onto a surface. Or hold ⌥ / Alt and drag the comp onto the surface.
  4. Let go. A wire now joins the comp to the surface, and the surface shows the comp. A notice says, for example, N ← wall-N.
  • One comp can feed many surfaces. Drag from its output dot to each surface.
  • If the surface already showed a comp, it now shows the new one.
  • To move a wire, drag its end onto another surface. You can also start the drag at the surface's input dot.
  • To remove a wire, drag its end onto empty graph. Or click the wire and press ⌫ / Backspace or Delete. The notice then says N unbound — no comp feeds it.
  • Click a wire to select it. Esc deselects it. Hover a wire: the tooltip names the comp and the surface, and says how to move it.
  • Each wire change is one undo step.
  • A surface keeps its own layers, and they draw in front of the comp. A surface with no comp and no layers is black.
  • A comp that feeds no surface sits below the surfaces, with no wire.
  • A small, dim fit: … label on a wire means that the comp's design size and the surface's size are different. The surface's fit setting maps one onto the other (Fit).
  • To place a comp as a layer on another surface, start from a layer that shows it (a use: layer, also marked ▸N). Right-click that layer and pick duplicate. Then hold ⌥ / Alt and drag the copy onto the other surface. The comp's own node at the top level has no duplicate item.
  • Select a comp to see it in Properties: its layers, their order and blends, and the surfaces that use it.
  • To give a comp a new name, see Rename a composition.
  • To give a comp a fixed size that crops it, for an AI input, see Comps with a frame.
  • Right-click a comp node for a colour and properties. Right-click a surface node for a colour and export that surface (Menus).

Rename a composition

Give a comp (a composition) a new name. Starlight changes the name everywhere that uses the comp.

In the node graph. Use the right-click menu. This is the only way to rename in the graph. A double-click on the node goes inside the comp (Step into a comp). No key starts a rename.

  1. Right-click the comp's node, or a node that uses the comp. A surface slice is a node that uses the comp, for example N · panorama.
  2. Pick Rename Composition…. On a node that uses the comp, the item names the comp and counts its uses, for example Rename composition 'panorama' (used by 5 places)….
  3. A text box opens on the node. Type the new name.
  4. Press ⏎ / Enter to save. Press Esc to cancel.

In Properties.

  1. Click the comp's node to select it.
  2. In Properties, type the new name in the name field.
  3. Press ⏎ / Enter, or click outside the field, to save.
  • The rename changes every place that uses the comp: each surface, each node that uses it, and each copy of it. You do not need to change anything else.
  • One comp can show on many surfaces. When you rename it from one surface, it changes on all the surfaces. The surfaces still show the same picture.
  • A rename is one undo step. Press ⌘Z / Ctrl+Z to get the old name back.
  • Starlight does not accept an empty name or a name that another comp has. A message says why, and nothing changes. A name also cannot contain / or :.
  • In the graph, a node that uses a comp on a surface shows the surface's name first, for example N · panorama and E1 · panorama. Then you can tell the surfaces apart. This is only the label in the graph. It does not change the scene file.
  • Claude can rename a comp too, with the rename_comp tool. Ask, for example: Rename the comp panorama to skyline.

Comps with a frame (crop for AI)

A comp can have a frame: a fixed size, as in After Effects. A framed comp is a box. It crops its content to that box. Use it to make an exact input for an AI generation, for example a 21:9 picture for Seedance.

The frame sizes.

  • AI shapes: the exact sizes Seedance makes. At 1080p: 16:9 1920×1080, 21:9 2206×946, 4:3 1664×1248, 1:1 1440×1440, 3:4 1248×1664 and 9:16 1080×1920. At 4K (Seedance 2.0 Pro only): 16:9 3840×2160 and 21:9 4398×1886. A frame of exactly this size is an input the model does not resize or crop again.
  • Surface: the size of one surface of the project, for example N 6400×1600 or STAGE 1920×1080.
  • Custom: type width x height in whole pixels, for example 2400x1200.

Make a framed comp.

  1. Right-click empty space in the node graph.
  2. Pick create › new comp…. Or pick Scene › New Comp… in the menu bar.
  3. Type a name. Pick a frame size.
  4. Click create. The new comp is placed in the middle of the surface or comp you clicked. It is selected, so its frame shows on the canvas.
  5. Double-click its node to go inside. Add layers there.

Pre-comp: put layers into a framed comp.

  1. Select the layers. They must be in the same list: the same surface, comp or group.
  2. Right-click one of them, on the canvas or in the graph. Pick pre-comp…. Or press ⌘/ / Ctrl+/ (or /) to open the command palette and run pre-comp…, or pick Scene › Pre-comp….
  3. Type a name. Pick a frame size.
  4. Click pre-comp.
  • The layers move into the new comp. One use: of the comp takes their place, in a transform.
  • Nothing moves on the canvas. The frame is centred on the layers.
  • The new comp turns around the middle of its frame, like every new layer (Move, scale, rotate). A new comp made with new comp… does too.
  • A part of a layer outside the frame is now cropped.
  • A pre-comp is one undo step.

See the crop.

  • Content outside the frame is not drawn. This is the same on the canvas and in an export.
  • The orange outline of the frame is a guide for you. It is not part of your content. It shows only on the canvas, never in an export or a render.
  • Select the comp: click it on the canvas, or click its node in the graph. A thin orange outline shows the frame's edge, with its name, size and shape, for example hero · 2520×1080 · 21:9.
  • Step into the comp (double-click its node). Now the frame is drawn strong, with the same label. The rest of the surface is dark.
  • At other times the canvas shows no outline. A surface that shows the comp as its feed (use: on the surface) shows the frame only when you are inside the comp.
  • To hide all frame outlines, choose View › Frame Guides. Choose it again to show them. The item has a ✓ while the guides show. Starlight remembers your choice. The guides are on when you first start Starlight. In the command palette, type frame guides.
  • Drag layers inside the frame. What you see inside the frame is what the AI model gets.
  • To move or scale the whole box on the surface, select the comp's node outside the comp and drag it. Its transform places the box.
  • The viewer shows the crop too (The viewer).

Change or remove the frame.

  1. Right-click the comp's node, a node that uses it, or a layer inside it.
  2. Pick set comp frame… (or comp frame… when it has one). Or pick Scene › Set Comp Frame….
  3. Pick a new size, or no frame. Click set frame. The content does not move.

Generate from a framed comp.

  1. Right-click the comp's node, a node that uses it, or a layer inside it.
  2. Pick generate from this comp…. Or pick Scene › Generate from Comp….
  3. Starlight renders the start frame from the comp. It saves the picture in assets/comps/<scene>/. The picture is an ordinary asset.
  4. The New Generation form opens. The picture is the start frame (for a video model) or the reference (for an image model). The model's aspect is set to match it.
  5. Read the line under the start frame. It says what Starlight does with your frame for this model.
  6. Type a prompt. Check the price. Click generate. Nothing is spent before you click it.
  7. When the clip is done, Starlight puts it back into the comp by itself (see below).
  • The picture shows the comp at the playhead.
  • The same picture is saved only once. Its file name has a short code for its pixels.

The shape of the clip. Models treat the shape of a start frame in different ways.

  • Seedance 2.5 keeps the shape of the start frame. A 2400×960 frame (2.5:1) makes a 2.5:1 clip. Nothing is cut off.
  • Seedance 2.0 (Pro, Fast and Mini on Magnific, and Seedance 2.0 on VideoRouter and Higgsfield) makes only six shapes: 16:9, 21:9, 4:3, 1:1, 3:4 and 9:16. It cuts any other start frame to its nearest shape, so the edges of your frame are lost. So Starlight makes the start frame bigger instead. Nothing of your frame is cut off.
  • Other models: Starlight sends the frame as it is, and sets the model's nearest shape. The model may cut the edges.

A bigger start frame (Seedance 2.0). An example: your frame is 2400×1200 (2:1).

  1. Starlight picks the Seedance shape that adds the least area to your frame. Here it is 16:9: the 16:9 box around your frame is 2400×1350.
  2. So Starlight adds 75 px above your frame and 75 px below it. The box is centred on your frame.
  3. Starlight renders this box at 1920×1080, the exact size Seedance makes at 16:9. So the model does not cut or resize it.
  4. The added band shows the comp's own layers. Layers often reach past the frame. In the room the frame hides them, but here the model sees them. So the model gets a normal picture of the scene.
  5. Where the band has nothing, Starlight stretches the edge pixels of your frame outward, softly. The band is never black.
  6. The form says this in one line: Seedance 2.0 makes 16:9. Starlight adds 75 px above and below from your comp, and crops them off when the clip comes back. Your frame gets about 89% of the clip's pixels.
  • Each resolution has its own sizes. At 1080p, 16:9 is 1920×1080. At 480p it is 864×496. At 4K it is 3840×2160.
  • When your frame already has a Seedance shape (for example 1920×1080 or 2206×946), Starlight adds nothing.
  • When you change the model or the resolution in the form, Starlight makes a new start frame for it. This takes a moment. A start frame made before is used again.
  • When you put a picture of your own in the start frame slot, the clip does not go back into the comp.

The clip goes back into the comp.

  1. When the clip is done, Starlight imports it into assets/. You do not need to click import.
  2. Starlight puts the clip into the comp, on top. Your own layers stay under it.
  3. The clip covers exactly the box of its start frame. The frame crops the added band away. So the clip lines up with your layers.
  4. This is one undo step. Press ⌘Z / Ctrl+Z to take the clip out again. Save the scene to keep it.
  • Starlight does this while the comp's scene is open. When another scene is open, the clip goes into the comp when you open its scene.
  • Each clip is put back only once. A retry makes a new clip, and it goes on top too.
  • A model can return a slightly different size. For example, Seedance 2.5 makes 1034×414 from a 2400×960 frame. Then Starlight scales the clip to fill the box. Less than one pixel is cut off.
  • When the comp is gone, or its frame changed, Starlight does not put the clip back. The clip is in the Gen panel and in assets/. Its card in the Gen panel says why.
  • A generation that Claude starts with render_comp_still and gen_create does not go back by itself. Place that clip by hand.

Scale to fit inside a framed comp.

  • Step into the framed comp. Right-click a layer and pick scale to fit (Scale to Fit).
  • Inside a framed comp, the frame is the box. Scale to Fit does not ask which surface. For example, a generated 21:9 picture inside a 21:9 comp fills the frame exactly with fill or width.

Ask Claude. Claude can do all of this with the precomp, set_comp_frame and render_comp_still tools. Ask, for example: Pre-comp the logo and the title into a 21:9 comp, then show me the frame.

In the scene file, a framed comp is written name: { size: 2520x1080, children: [ … ] }. A comp without a frame is a plain list, as before. A bypassed comp is written name: { children: [ … ], bypass: true }, and a framed one keeps its size (Bypass a whole comp). A scene with a bypassed comp needs Starlight 0.2.14 or newer.

Words. Start frame: the picture a video begins with. Band: the strip Starlight adds around your frame to make the start frame bigger.

Step into a comp

Work on the layers inside a comp.

  1. Double-click a comp's node (any node with ▸N), or a surface slice that uses a comp. You can double-click anywhere on the node. The name works too. To rename a comp, right-click its node (Rename a composition).
  2. The graph now shows only that comp, and it fits the view. The breadcrumb at the top left shows where you are, for example scene ▸ comp.
  3. To step back out, click a name in the breadcrumb. Click the scene name to go back to the top level (its tooltip says Back to the top level).
  • Comps inside comps open in the same way.
  • The breadcrumb shows only while you are inside a comp.
  • Each scene remembers the level you were at.
  • To find a node at any level, press ⌘F / Ctrl+F (Find a node).
  • The level also decides what a click on the canvas selects. At the top level, a comp on a surface is one layer. Step into the comp to click its own layers on the canvas (Select a layer).
  • A double-click on the comp's art on the canvas also steps into the comp. Then a click on another surface or on empty canvas, or Esc, steps back out to where the double-click started.
  • When you change level, nodes that the new level does not show leave the selection. A ⌘F / Ctrl+F jump keeps the node it found.

Find any node in the scene by its name, at any level, and jump to it.

  1. Press ⌘F / Ctrl+F. A search box opens at the top. Its grey text says find a node — any level….
  2. Type part of the node's name.
  3. Use ↑ ↓ and ⏎ / Enter to pick one, or click it.
  • When you pick a node, the graph steps into its comp, selects the node and shows it.
  • Each row also shows where the node lives: ⊂ and a comp name, or ▸ and a surface name. You can type that part too.
  • With no match, the list says no node matches. The list shows up to 60 nodes.
  • To close the box, press Esc or ⌘F / Ctrl+F again, or click outside it.
  • With the pointer over Properties, ⌘F / Ctrl+F filters the fields in Properties instead.

Wire and reorder layers

Move a layer into another parent, change the layer order, or switch nodes off.

Move a layer under another parent.

  1. Hold ⌥ / Alt. Over a node, the pointer shows a link arrow.
  2. Drag the node. A wire follows the pointer, and the places where you can drop light up.
  3. Let go on a transform, a group, a comp's root or a surface.
  • To plug a node into a matte node, drop it on the top half (the content) or the bottom half (the matte). What was in that input before becomes a layer just above the matte node.
  • An effect chip moves with its node: an ⌥ / Alt-drag on a chip moves the node it belongs to.
  • Surfaces, comps and clock dials cannot be moved this way.
  • A node cannot go inside itself. If you let go anywhere else, nothing changes.
  • A short ⌥ / Alt-click (under 6 px) adds the node to the selection or takes it out, as usual.
  • Each move is one undo step.

Change the layer order. Select a layer, then press:

⌘] / Ctrl+]One step forward (up the stack)
⌘[ / Ctrl+[One step back (down the stack)
⇧⌘] / Ctrl+Shift+]To the front
⇧⌘[ / Ctrl+Shift+[To the back

These keys move only the main selected layer (the last one you selected). You can also select the comp and use ↑ and ↓ in its layers list in Properties (Properties).

Switch a node off, or delete it.

  • Press b to bypass the selected nodes, or click the small round switch on a node's left edge. A bypassed node is dim, with a hollow outline. Bypass switches off what the node adds (Turn a node off). Press b again to switch it back on.
  • With several nodes selected, b gives them all the same state. If at least half of them are on, all switch off. If most are off, all switch on.
  • Press ⌫ / Backspace or Delete to delete the selected nodes. A surface cannot be deleted.
  • To compare a look with and without one effect, click that effect's chip, for example glow 0.25. Then press b. Only that effect switches off.
  • To switch a whole chain of effects, select its chips with a box, then press b.
  • With a chip selected, ⌫ / Backspace removes only that effect.

The node menu. Right-click a node for these items. The colour swatches are at the top (Colour and note nodes).

ItemDoes
propertiesShows the node in Properties (⇧P / Shift+P)
open in timelineShows the node's animation in the timeline. Only for animated nodes
promote to named clockMoves the node's own timing to a named clock that other nodes can share. Only for nodes with their own animation
bypassSwitches the node off or on (b). A dot (●) shows when it is off
reveal source in finderOpens the folder of the image, video or cutout file, with the file selected. On Windows the item says show source in explorer
duplicateCopies the node, with its transform. The copy goes on top, a little down and to the right, and Starlight selects it
group into subtreePuts the selected layers into a new comp (Make a comp)
pre-comp…Puts the selected layers into a new comp with a frame, and nothing moves (Comps with a frame)
make loop…On a video layer: makes the clip loop with no jump (Make a clip loop)
bring forward, send backward, bring to front, send to backChanges the layer order
add siblingAdds a new layer next to this one
deleteDeletes the node
  • Right-click a surface to export only that surface.
  • Right-click an effect chip for three items: properties, bypass with the effect's name (for example bypass glow), and remove effect.
  • Every menu in the graph: Menus.

Turn a node off (bypass)

Bypass switches off what the node adds. The node stays in the graph, so you can switch it back on at any time.

  1. Find the small round switch on the left edge of the node. It is faint while the node is on.
  2. Click the switch. The node turns off at once. The switch is bright, and the node is dim.
  3. Click the switch again to turn the node back on.
  • A click on the switch does not select the node, and it does not move it.
  • If the node is part of the selection, the click turns off (or on) every selected node. This works like b. It is one undo step.
  • The tooltip on the switch says bypass (b). With a node selected, the key b does the same.
  • When you zoom the graph far out, the switches hide, like the other small marks.

What bypass does to each kind of node.

NodeBypassed, it
transform, groupPasses its layers through. They draw and play as if the transform or group was not there: none of its position, scale, rotation, anchor, skew, opacity, blend, animation, gate, in/out, delay or effects
a picture, video, shape, gradient, text or pathHides. Nothing is under it to pass through
use of a comp, cloneHides, with all its copies
audioIs silent
matteShows its content without the cut
adjustment layerChanges nothing. The layers below it draw as they are
an effect chipTurns off only that effect (Turn one effect off)

A bypassed video costs almost nothing. It does not play with the scene, it does not change to its full-size file when you zoom in, and an export or a still does not open it. So you can keep old takes in the scene as bypassed layers. When you switch one back on, it shows its frame after a short moment.

Bypass a whole comp. Select the comp's root node (or a surface's comp node), then press b or click its switch. The comp turns off everywhere it is used: every use of it, every clone of it, and every surface that shows it. Those nodes show dim. A surface that shows the comp still draws its own layers. To turn off only one place where the comp is used, bypass that use node instead.

Colour and note nodes

Colours and sticky notes help you organise a busy graph. They never show on the surfaces or in an export.

Colour a node.

  1. Right-click the node.
  2. Click a swatch in the row at the top of the menu: red, orange, yellow, green, blue or purple.
  3. To clear the colour, click the ringed swatch again, or click none.

Add a sticky note.

  1. Right-click empty graph where you want the note.
  2. Pick add note.
  3. Type your text.
  4. To finish, press Esc or ⌘⏎ / Ctrl+Enter, or click outside the note.
  • Layers, comps, surfaces and clock dials can all have a colour. Hover a swatch to read its name.
  • Right-click a node in the selection to colour every selected node.
  • The colour shows as a light tint. A comp's colour shows everywhere the comp appears. An effect chip has the colour of its node. A layer's bar in the timeline and the keyframe ticks of its ring on a clock dial have its colour too.
  • The colours have no fixed meaning. They are not the grades in Generations.
  • An empty note shows the grey word note. When you zoom out, the text stays readable.
  • To edit a note, double-click it. Or select it and press ⏎ / Enter.
  • While you type in a note, Starlight's shortcuts are off. ⌘Z / Ctrl+Z undoes your typing.
  • Drag a note to move it. Drag its bottom-right corner to resize it.
  • To delete a note, select it and press ⌫ / Backspace or Delete. Or right-click it and pick delete note.
  • Right-click a note for its own menu: the colour swatches, edit note and delete note.
  • To deselect a note, press Esc, or click outside the graph.
  • A note shows only where you made it: at the top level, or inside one comp. Notes sit behind wires and nodes, and they never block a click.
  • Colours and notes are saved with the scene, and each change is one undo step. Claude can read the notes.
  • Claude can also write a note. A note from Claude, or from the scene file, may have no place on the graph yet. Starlight puts it in an empty spot next to the node it is about (or next to the first column), and never on top of a node or another note. If there is no room near, it goes below all the nodes. The note is as big as its text.
  • When you move or resize such a note, it stays where you put it. Starlight never moves a note that you placed.
  • Fit (double-click empty graph) shows the notes too.

The viewer toggle v

The viewer shows one node's own result on its surface, like the viewer in Nuke. Use it to check one step of a look.

  1. Select a node in the graph.
  2. Press v, or click viewer at the graph's top right, or pick View › Viewer. The menu item has a ✓ while the viewer is on.
  3. The surface that the node feeds now shows the node's result, not the final output. Other surfaces keep their output.
  4. Press v again to see the output on every surface.
Not the final output

A surface in the viewer has a thin grey frame just outside it. The viewer button and the viewed nodes turn magenta. Export, Glimpse and Claude's renders always use the output, never the viewer.

What the surface shows.

You selectThe surface shows
A layerThe layer by itself, still, at its own origin. Nothing downstream of it applies (see below)
A transformIts layers, with its own position and animation
An effect chipThe node with its effects up to this one. Click along a chain to see each step of the look
A matte nodeThe cut result
A matte node's content inputThe content, not cut
A matte node's matte inputThe mask itself, white on black
An adjustment layerThe layers below it, adjusted
A compThe comp alone, on every surface that uses it, in the place where the surface shows it. The surface's own layers do not show
A node inside a compOnly this node's part of the comp, in the place where the surface shows the comp
Several nodesEvery selected node that reaches the surface (see below)
Nothing, a surface, an audio node or a clockThe normal output. A line under the viewer button says why
  • Downstream means everything after the node: its own effects and blend, and the position, scale, animation, effects, opacity and blend of its parents in the same comp or surface. It also includes a matte's cut and an adjustment layer above the node.
  • Inside a comp that several surfaces share, each surface shows its own part of the comp, where that part appears in the room.
  • The surface keeps the position, scale and rotation that it gives the comp, and their animation. Its fit stays the same too. Its effects, opacity, blend and adjustment layers do not show.
  • If the node is not in a surface's part of the comp, that surface is empty. This is correct: the node is not on that surface.
  • To view several nodes, select them all, for example with a box. Each surface shows every selected node it reaches, stacked in the comp's own order.
  • The backmost node draws normally. Each node in front of it blends over it as it does in the comp.
  • A node and its parent show as the parent. Nodes on different surfaces each show on their own surface.
  • If two selected nodes are on the same surface but in different comps, the surface shows only the one you selected last.
  • The line under the button says, for example, select a node to view it, or scene clock: a clock has no image · showing the output.
  • To move time while you view a node, drag sideways on a clock dial's ring, drag on an audio or video dial, scroll over a dial, click ‹ or › on the scene dial, or use the timeline. The selection stays the same, so the surface still shows only the selected nodes, at the new time. A click on the ring of a clock dial (with no drag) selects the clock, and the surface then shows the output. A click on the face of a video dial selects the clip, and the surface then shows the clip. A click in the centre of any dial only plays or pauses: the selection stays, so the surface does not change.
  • The button's tooltip says viewer (v) · show the selected node's result on the canvas. While the viewer is on, the tooltip starts with viewer on (v).
  • The command palette has it too: toggle viewer.
  • The viewer is off each time you start Starlight or open a project.

Use the clock dials

A clock sets the timing of the animation that uses it (Clocks). Each clock is a round dial in the lane on the right of the graph.

  • The large dial at the far right is the scene clock. Its name, scene, is above it.
  • Named clocks stack to its left, each with its name above, for example half. Each one is wired into the scene clock and out to every node that uses it.
  • The centre of a dial shows the time so far at the top. A play or pause sign is in the middle. The clock's length is at the bottom, for example 2.00s. A hand sweeps round as the clock runs.
  • A keyframe is a value at one moment. Each animated layer gets one ring on the dial of the clock it uses. The ring shows all the layer's keyframes: its motion, its colours, its gradient and its effects. Each keyframe is a tick. 12 o'clock is the start, and time runs clockwise (Read the keyframe rings).
  • A layer's on and off times (its gate) show on its ring. The ring is green where the layer is on. Edit them in the timeline.
  • A cue is a named moment on a clock. A trigger makes another clock start at that cue (Triggers).
On a dialDoes
Drag sideways on the ringScrubs time. Hold ⇧ / Shift for 4× finer steps. Scrubbing pauses playback
Scroll over the dialMoves the playhead one frame for each click of the mouse wheel. Hold ⇧ / Shift for 10 frames. Scrolling pauses playback (Scrub to a time)
Click the centrePlays or pauses. It does not change the selection. The sign in the centre shows what a click does: ▶ while the scene is paused, ⏸ while it plays
Click ‹ or › beside the scene dial's play controlSteps one frame back or forward. Hold ⇧ / Shift for 10 frames. Hold the button to repeat. The same as , and .
Click the time in the centreSwitches the whole app between seconds and frames
Hover a ringMakes that ring and its ticks brighter (a coloured layer's ticks show their full colour), and the other rings go dim. The centre shows the ring's layer, for example shape rect. The ring of the selected layer stays a little brighter
Hover a keyframe tickShows what it is, and where, in the centre. For example fill 50%, or 3 keys 25% for a tick that holds three keyframes
Click a ring or a tickSelects that ring's layer. A tick of one motion keyframe also shows that keyframe in Properties
Drag a tick along its ringMoves its keyframes in time, together. They cannot pass the keyframes beside them. Press Esc to cancel
Click the empty cue band (it shows when you hover the dial)Adds a cue there
Drag a cue along the bandMoves the cue in time. The centre shows its frame
Drag a cue off the band onto another dialMakes that clock start at this cue (a trigger). Only a named clock can be the target
Drag the dial's nameMoves the dial in the graph
Drag the small ↗↙ handle beside the name sceneMakes the scene clock bigger or smaller. Double-click the handle for the normal size (Make the scene clock bigger)
Right-click the dialOn the scene dial: the transport buttons first (see below). Then properties, rename, select users, delete, add cue at playhead, clear cues and clear trigger

Right-click the scene dial for its transport buttons. They are at the top of the menu, next to the pointer. The buttons that you use most are nearest.

RowButtons, from left to right
1play or pause (Space: ▶ while paused, ⏸ while playing) · cache and play the loop (⇧Space / Shift+Space, the RAM preview) · go to first frame (Home) · play from first frame
2previous frame (,) · next frame (.) · previous keyframe (j) · next keyframe (k)
3go to last frame (End) · previous marker · next marker · work area in / out
  1. Right-click the scene dial.
  2. Click a button. The menu stays open, so you can click again. For example, click next frame many times to step through the scene.
  3. Press Esc, or click outside the menu, to close it.
  • Point at a button to see its name and its key. A button does the same thing as its key.
  • A button is grey when it has nowhere to go. For example, next keyframe is grey after the last keyframe.
  • previous keyframe and next keyframe go to the keyframes of the selected layers. When no selected layer has keyframes, they go to every keyframe in the scene. A right-click on the scene dial selects the scene clock, so from the dial they use every keyframe. They are the same as j and k (Go to the previous or next keyframe).
  • previous marker and next marker go to the timeline markers (Add timeline markers).
  • play from first frame goes to the first frame and plays. With a work area set, playback starts at the start of the work area.
  • The work area button has three steps. Its first click sets the start at the playhead, as i does. Move the playhead, and the second click sets the end, as o does. The third click clears the work area. The button's tooltip says what the next click does (Loop a part of the scene).
  • The work area button is lit while a work area is set.
  • The same buttons show when you right-click ▶ or the time at the bottom left of the canvas.
  • The command palette has the same commands, for example next keyframe or play from first frame.
  • A click on a dial's ring, or on its name, also selects its clock. A click in the centre does not: it only plays or pauses. A drag on the ring does not select either. The selection stays, so the viewer keeps showing the selected nodes while you play or scrub.
  • With the pointer over a dial, ← and → step one frame (10 with ⇧ / Shift).
  • The scene clock has only properties and the cue items. You can delete a named clock only when no node uses it. Until then, delete is grey and says, for example, used by 2 nodes.
  • select users lists the nodes that use the clock. Pick one to select it.
  • clear cues shows only when the clock has cues, with their number in brackets. clear trigger shows only on a clock that a trigger starts.
  • When a dial has more animated layers than rings, the last ring is +n more. Its +n mark says how many layers share it (Read the keyframe rings).
  • An amber ring means that the clock's length does not fit a whole number of times into the scene's loop. The animation then jumps where the export loops.
  • When you zoom out, the scene dial keeps its full size and draws over its neighbours. Its controls stay easy to click.
  • A scene dial that you made bigger gets smaller when you zoom out, but never smaller than its normal size (Make the scene clock bigger).
  • While you drag a ◇ from Properties, the dials light up as places to drop (Animate a property on a clock).

Read the keyframe rings

The rings around a clock dial show when things change in the loop. There is one ring for each animated layer that uses the clock.

  • A ring shows every keyframe of its layer on this clock: position, scale, rotation and opacity, the gate, the fill and stroke colours of a shape, a gradient's colours and other values, and the values of its effects, for example a glow.
  • Each keyframe is a small tick across the ring. 12 o'clock is the start of the loop, and time runs clockwise. The hand shows where the playhead is.
  • The ticks have the colour of their layer: the colour you gave the layer in the graph (Colour and note nodes). It is the same colour as the layer's bar in the timeline. The ticks are a soft version of the colour. They show the full colour when you point at the ring. A layer with no colour has grey ticks. The +n more ring is always grey, because it holds many layers.
  • Keyframes that are very close together share one tick. A longer, heavier tick holds two or more keyframes.
  • Many keyframes in a row show as a solid line along the ring. For example, tracking data with a value on every frame.
  • A layer with a gate has a dim ring. The ring is green where the layer is on.
  • The rings are in the same order as the layers in the scene, from the inside out. They do not move while the scene plays.

When there are many layers. A dial has room for only some rings. At its normal size, a dial shows five rings. The last ring is then +n more: it holds the keyframes of all the other layers, and its line is dashed. The +n at the top says how many layers share it.

To see more layers on their own rings:

  1. Select a layer. A selected layer always has its own ring. If it was in +n more, it takes the last ring before +n more.
  2. Make the dial bigger. Zoom in on the graph, or make the scene clock bigger (Make the scene clock bigger). The rings stay at least 4.5 pixels apart, so a bigger dial has room for more of them: about 9 at twice the size, 17 at four times, and up to 24.
  • Point at a ring to see its layer's name in the centre of the dial. Wait a moment to see a tooltip with the layer, its animated values and its number of keyframes. On +n more, the tooltip lists the layers.
  • Click a ring to select its layer. Click a tick on +n more to select the layers that it holds. They then get their own rings.
  • Drag a tick to move its keyframes in time. You cannot drag a solid line or a tick on +n more. Select the layer first.
  • A dial draws only a small number of marks, however many keyframes the scene has. So a big scene stays fast.

Make the scene clock bigger

When a scene has many keyframes, their ticks are close together on the scene clock's rings. Make the scene clock bigger to give them more room. A bigger scene clock also has room for more rings, so more layers get a ring of their own (Read the keyframe rings).

  1. Point at the name scene above the scene clock. A small ↗↙ handle shows to the right of the name.
  2. Drag the handle up and to the right. The scene clock gets bigger while you drag.
  3. Release the mouse button. Starlight keeps this size for the project.
  • To make it smaller, drag the handle down and to the left.
  • To go back to the normal size, double-click the handle.
  • To cancel a drag, press Esc before you release. The scene clock goes back to the size it had.
  • The handle's tooltip says drag to resize the scene clock · double-click: default size.
  • A drag on the handle only changes the size. It does not select the clock, move time or move the dial.
  • The rings, the keyframe ticks and the hand get bigger, and more rings fit. The text and the buttons in the middle stay the same size.
  • The scene clock can be up to 8 times its normal size. It can never be bigger than the graph area at 100% zoom. If you make the graph area smaller, the scene clock gets smaller too. If you make it bigger again, the scene clock gets its size back.
  • The graph makes room for a big scene clock. Nodes do not go under it, and ⇧1 / Shift+1 in the graph fits it on screen.
  • The size is saved with the project's layout, not in the scene. It is not an undo step, and it does not make the scene unsaved. Another project, or a computer without this project's layout file, uses the normal size.

Use the audio and video dials

An audio or video layer shows as a round dial in the graph. Use it to see where you are in the sound or the clip, and to move through it.

The audio dial.

  • The top line shows ♫ and the layer's name.
  • The outer ring fills as the track plays, from start to end.
  • The centre shows the tempo, for example 120 bpm, and the time, for example 0:42 / 3:10. Until Starlight has analysed the song, the centre shows ♫. Until it knows the length, the time shows ….
  • After the analysis, two thin rings fill with each bar and each beat.
  • Coloured arcs show the parts of the song. Parts that repeat, such as every chorus, share a colour. Hover an arc to see its part letter and its energy. Dots mark the drops, and a larger dot marks the high point.

The video dial.

  • The top line shows ▸ and the layer's name.
  • The ring fills as the clip plays.
  • The centre shows the clip's time. While the scene is paused, a ▶ is in front: a click on the centre plays. The clip's length is below. Until the clip is ready, it says decoding….
  • Video layers follow the scene's playback (Transport).
On an audio or video dialDoes
Drag sideways on the faceScrubs through the track or the clip. A drag of about 480 px covers the whole length. Hold ⇧ / Shift for 4× finer steps. Scrubbing pauses playback
Scroll over the dialMoves the scene one frame for each click of the mouse wheel. The track or the clip follows. Hold ⇧ / Shift for 10 frames (Scrub to a time)
Click the centrePlays or pauses the whole scene. The clip follows. It does not change the selection
Drag the nameMoves the dial in the graph
  • A click on the face (with no drag) also selects the layer. A click in the centre does not: it only plays or pauses.
  • A drag on the face does not change the selection. So the viewer keeps showing the selected nodes while you scrub.

Animate a property on a clock

Connect a property to a clock dial, so that the clock drives the property's animation.

  1. Select a layer. In Properties, find the property, for example its position.
  2. Drag the small ◇ beside the property onto a clock dial. The dials light up while you drag.
  3. Let go on the dial. The property is now animated on that clock. A notice says, for example, animating x on clock "beat".
  • On a named dial, the property shares that clock's timing with other nodes. On the scene dial, the node gets its own loop.
  • The timeline opens on the new lane, and Properties shows the layer.
  • If the property is already animated, nothing changes. The notice says, for example, x is already animated.
  • Press Esc while you drag to cancel.
  • A plain click on ◇ starts the animation on the node's own clock. Then add keyframes (Animate a property).
  • On a property that already has keyframes, a click adds a keyframe at the playhead, or removes the one under it.
  • Effect settings have the same ◇ chip.

Find the hidden gestures

Four gestures have nothing on screen that shows them. The graph gives a small hint for each one.

GestureHow
Select with a box (marquee)Right-drag across empty graph
Move a node under another parent (re-parent)Hold ⌥ / Alt and drag the node onto a transform, group, comp or surface
Start a clock at a cue (trigger)Drag a cue off its band onto another dial
Switch seconds and framesClick the time in a dial's centre
  • The hint is a dim line in the graph's bottom-left corner. It shows while the pointer is where the gesture starts, or while you do the gesture.
  • For example, over empty graph with something selected, it says right-drag to select several · ⇧ adds to the set · ctrl-click adds or removes.
  • During an ⌥ / Alt-drag, it says drop on a lit group, comp, surface or matte port · anywhere else cancels.
  • After you use a gesture twice, its hint stops. Starlight counts the uses on each computer separately.
  • To see the hints again, open the command palette and pick show graph hints again.
  • The palette lists the four gestures too: marquee select, re-parent a node, wire a cue trigger and seconds ⇄ frames. Type marquee, trigger, re-parent or frames to find them.
  • Every key and mouse action for the graph is in Keyboard & mouse › Node graph.

On this page