Graphics

Kinema's graphics work like Premiere's Essential Graphics: a graphic clip on a video track holds text and shape layers. Text is set by the text engine in crates/text (README): bundled OFL fonts plus system fonts, OpenType shaping with kerning and ligatures, bidi, line breaking, all drawn as vectors in linear light.

Model

• A graphic clip's project item is an ItemKind::Graphic { width, height, rate }: a transparent canvas the size of the sequence frame, with unlimited duration. These items are internal and do not appear in the Project panel.
• Its layers are effect instances graphic_text / graphic_shape in the clip's effect list, in paint order (first = back). They are hidden from the Effects panel. Because they are effect instances they save with the project, take keyframes on every animatable property (Source Text keyframes hold), appear in Effect Controls and copy with the clip. The Essential Graphics panel's number and point properties (Tracking, Leading, Font Size, Position, Opacity…) carry the same stopwatch as Effect Controls (graphics.prop.<id>.stopwatch, #2358): once a property animates, changing it in the panel adds or updates a keyframe at the playhead. Standard effects on a graphic clip apply to the composed graphic; Motion and Opacity apply last.
• kinema_project::graphic::eval_layer evaluates a layer at a time into a LayerSpec.
• Projects with graphics need no schema migration: everything is additive (new item kind, new effect ids). Builds older than this one refuse such files with a parse error.

Layer properties

Point text's origin is its alignment point on the first baseline (the click point of the Type tool); paragraph text's origin is the top-left corner of its box; a shape's origin is its centre. Colours are #rrggbb (sRGB).

Point text and paragraph text

A text layer is one of two types, as in Premiere Pro (behaviour observed in Premiere Pro 26):

• Point-text scale. Only the pointer's travel along one side of the box counts: along its width for the two side handles, along its height for the corners and the top and bottom handles. Travel d away from the anchor scales by 1 + d / r, where r is the handle's distance from the anchor along that side, plus 60 points when the handle is nearer than 60 points (so a handle on the anchor does not scale without bound). A handle exactly on the anchor grows the text when dragged into the box.
• Paragraph box. A side dragged past its opposite stops at half the font size. When the top or left side moves, the layer's origin moves with the corner, so the anchor point is renumbered to stay on the same spot of the picture and Position does not change.
• Anchor point. Drag it with the Selection tool: the anchor moves, the layer stays (Position and Anchor Point change together). It is picked before anything under it: a shape dragged by its exact centre moves its anchor, not the shape, and the top-left handle of a new paragraph box is reached only after moving the anchor.
• Switching type. Properties ▸ Text ▸ wrench ▸ Text Properties ▸ Text Layer Type (graphics.setTextType). To point text: the wrapped lines become real lines (also those the box hid). To paragraph text: the box is fitted to the text. Either way the text stays where it is and Position is untouched; only Anchor Point is renumbered. Vertical text is always point text.

Known differences from Premiere: the first line of paragraph text sits one font ascent below the box's top (Premiere places it higher, about one cap height below); Premiere's smallest box and the box it fits when switching to paragraph text were not measured exactly; the 60 points were measured with the Program monitor at Fit only; Text Properties has no Hindi Digits option.

Rendering

crates/render/src/graphic_clip.rs rasterises each layer straight to the output transform (sequence Motion × layer transform), so text stays sharp at any scale or rotation. Fill coverage comes from the text engine; strokes come from a Euclidean distance transform of that coverage (outer strokes sit under the fill, centre and inner strokes over it); the shadow is the union of all parts, spread by Size, offset and blurred. Each layer's raster is cached while it stays the same. The GPU plan receives one tight image per graphic clip and places it with a translation. The Timecode and Clip Name effects also draw with the text engine.

Commands

Without clip, commands use the selected graphic clip, else the topmost graphic clip under the playhead. Without layer, they use the selected layer, else the front one. New graphic clips are placed at the playhead on the first video track above the clips there that is free for the duration (a track is added if needed).

Program monitor

Moving a layer snaps its edges or centre to the frame edges, the frame centre and the guides (View ▸ Snap in Program Monitor, on by default; hold ⌘/Ctrl to move freely). See monitors.md for rulers and guides.

Automation ids: program.layer.<clip>.<layer>, program.layer.<clip>.<layer>.handle.<n> (0–3 the corners from the top-left clockwise, 4–7 the top, right, bottom and left sides), program.layer.<clip>.<layer>.anchor, program.textEdit (while editing).

Properties / Essential Graphics panels

With a graphic clip selected, the Properties panel (and Essential Graphics) shows: Layers (front first; new text / rectangle / ellipse, bring forward / send backward, delete, visibility), Align and Transform (six align buttons, two distribute buttons, position, anchor, scale, rotation, opacity), Text (Source Text field, font family and style, size, paragraph alignment, tracking, leading, baseline shift, box width and height, faux bold / faux italic / all caps / small caps / underline, kerning and ligatures; the wrench in the section's header opens Text Properties: Text Layer Type and Ligatures), Shape, and Appearance (fill, two strokes, background, shadow). Automation ids: graphics.layers.<n>, graphics.prop.<parameter id>, graphics.align.<how>, graphics.distribute.<axis>, graphics.sourceText, graphics.NewTextLayer, graphics.NewRectangle, graphics.NewEllipse, graphics.deleteLayer, graphics.section.<name>, graphics.textProperties (the wrench) and in its dialog graphics.textProperties.type, .type.point, .type.paragraph, .ligatures, .ok, .cancel.

Per-character styles

Select characters with the Type tool and change font, style, size, colour, faux bold / italic, underline, tracking, baseline shift or caps in the Properties panel: the change applies to the selection only (graphics.setCharStyle). Styles are stored on the layer as runs of characters (EffectInstance::layer.runs, character offsets, sorted and non-overlapping) and follow the text when it is edited (typed characters take the style of the character before them). The text engine lays the runs out together (kinema_text::layout_rich): a run can change the font and size, lines grow to fit the largest run, and each run can have its own fill colour.

Responsive design

Position (pins). A layer's edges can be pinned to another layer of the same graphic or to the video frame (graphics.pin {layer, to: "frame" | layer index | name | "none", edges}). Each pinned edge keeps its distance to the same edge of the target, measured when the pin is made: a box pinned on all four edges to a text layer grows with the text; a shape pinned on two opposite edges is resized, a text layer only moves. Pins resolve in dependency order (cycles are ignored). Moving a pinned layer keeps the pin at its new distance. Layers get a stable uid for this.

Time (intro / outro). graphics.setResponsiveTime {introFrames, outroFrames} protects the first and last part of a graphic: when the clip is trimmed or extended, keyframes inside the intro keep their timing from the clip start, those inside the outro from the clip end, and the middle is stretched or squeezed (graphic_design::remap_time). Trimming the head keeps the intro playing from the new start.

Rolls and crawls. graphics.setRoll {mode: off | roll | crawlLeft | crawlRight, startOffScreen, endOffScreen, prerollFrames, easeInFrames, easeOutFrames, postrollFrames} (the Roll options of Responsive Design – Time) moves the whole graphic: up for a roll, sideways for a crawl. Start Off Screen starts with the content just outside the frame (below it / right of it / left of it), otherwise where it was laid out; End Off Screen lets it leave completely, otherwise it stops when its last edge reaches the frame edge. The speed is constant between the preroll and the postroll, with linear acceleration over Ease In and deceleration over Ease Out (graphic_design::roll_progress). The offset is exact maths on the union of the layers' bounds (with background padding), at the clip-relative time.

These settings are TrackItem::graphic (GraphicMeta, schema v12).

Graphics templates (.fcgt)

Kinema's equivalent of motion graphics templates is its own format: a .fcgt file is documented, versioned JSON (kinema_project::gtemplate):

Placing a template on a sequence of another size scales and centres its layers (pin distances scale too).

Kinema never reads or imports Adobe .mogrt files (or any other application's template packages): Install refuses .mogrt paths without opening them, and the reader rejects ZIP archives. Built-in templates are original designs made for Kinema in code (no image, font or template files): lower thirds (Slab, Rule, Ticker crawl), titles (Centered, Boxed), end credits (Roll) and callouts (Pointer, Tag), with the bundled Inter font. Thumbnails are rendered by the engine at runtime.

User templates live in <data dir>/Graphics Templates/.

Essential Graphics panel. Browse lists the built-in and user templates as cards with engine-rendered thumbnails, a search field and a category filter; click selects, double-click or Apply places the template at the playhead, dragging a card onto a video track places it there, Install… picks a .fcgt, Remove deletes a user template. Edit is the graphic editor; for a graphic made from a template it starts with Template Properties (one editor per control). With no layer selected it shows Responsive Design – Time (intro / outro, Roll with its options); with a layer selected, Responsive Design – Position (Pin To and the pinned edges). With characters selected by the Type tool, the Text section's controls style just those characters. Export As Motion Graphics Template… opens a dialog (name, category, description and a checklist of properties to expose); Replace Fonts in Projects… lists the fonts in use (missing ones flagged) and replaces one. Automation ids: essentialGraphics.tab.browse|edit, gfxTemplates.search, gfxTemplates.category, gfxTemplates.item.<id>, gfxTemplates.apply|install|remove, gfxTemplates.control.<control id>, graphics.roll.*, graphics.time.intro|outro, graphics.pin.to|left|top|right|bottom, exportTemplate.*, replaceFonts.*; ui.set {"gfxTemplates": {...}, "gfxEdit": {...}} sets the panel state and the Type-tool selection.

Captions, source graphics and fonts

Gradient fills

Appearance › Fill Type › Linear Gradient paints the shape or the text with a ramp from Gradient Start to Gradient End. Stops are blended in linear light. The angle is degrees clockwise on screen; 0° runs left to right across the layer's own bounds, and the ramp turns with the layer. graphics.set accepts fill_kind (solid or linear gradient), gradient_start, gradient_end and gradient_angle. A per-character fill colour still paints those characters as a solid; the rest of the text keeps the ramp. A singular layer transform falls back to the solid fill colour.

Not yet

Mask-with-text, per-layer blend modes, Bézier curves in the pen tool (paths are polygons), media layers inside graphics, and template controls grouped into folders.