@vosjs/cli 0.55.0 → 0.55.1

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@vosjs/cli",
-  "version": "0.55.0",
+  "version": "0.55.1",
   "description": "The vos CLI: record the real product from a scripted browser flow, auto-zoom from the cursor track, cut as data in doc.json, render deterministic video and stills, deliver a release's media per destination spec, and sync with vos.so. One binary, every verb, MIT.",
   "type": "module",
   "license": "MIT",
@@ -51,11 +51,11 @@
     "acorn-walk": "^8.3.5",
     "mediabunny": "^1.55.7",
     "playwright": "^1.49.0",
-    "@vosjs/core": "^0.25.5",
     "@vosjs/editor": "^1.4.0",
-    "@vosjs/elements": "^0.9.0",
-    "@vosjs/render-core": "^0.3.11",
+    "@vosjs/elements": "^0.10.0",
+    "@vosjs/core": "^0.25.5",
     "@vosjs/shared": "^0.6.0",
+    "@vosjs/render-core": "^0.3.11",
     "@vosjs/studio-core": "^0.33.2",
     "@vosjs/timeline": "^0.4.1",
     "@vosjs/tween": "^0.8.2"

package/skills/VERSION

@@ -1 +1 @@
-0.14.0
+0.15.0

package/skills/vos-authoring/SKILL.md

@@ -164,6 +164,37 @@ Configs pushed to vos.so get preview-rendered on a software-GL fleet: no
 (too slow), and any fetched asset must be reachable and CORS-open. These are
 fine for purely local renders on a real GPU.
 
+## Your own passes: depth of field, refraction
+
+The `postprocessing` list has no depth-of-field pass. For lens effects,
+render the world yourself:
+
+- Build the world in a private `THREE.Scene` with its own camera. In the
+  timeline's `onUpdate`, render it into `WebGLRenderTarget`s (`samples: 4`
+  plus a `DepthTexture` works, the render fleet included), run your passes,
+  then restore the renderer's target, clear colour and `autoClear`.
+- `ctx.scene` holds ONE full-screen quad that composites the result
+  (`gl_Position = vec4(position.xy, 0.0, 1.0)`, `frustumCulled = false`, no
+  depth test). Add it with `ctx.scene.add()`: the engine never adds
+  `objects` for you, and a quad left out renders a flat black frame.
+- Size the targets from `renderer.getDrawingBufferSize()`; rebuild them
+  when it changes.
+
+Depth of field that reads:
+
+- **Set the blur in pixels, by design.** A physical lens at scene scale
+  (f/1.6 at 40 cm) leaves a sharp zone a few millimetres deep, so the
+  subject you mean to show goes soft too. Decide what one depth step should
+  cost (8 to 20 px at 1080p) and solve for the lens strength.
+- **Keep the subject square to the camera while it is in focus.** A plane
+  seen at an angle spans depth, and shallow focus sharpens only a strip of
+  it. Check that the blur at the subject's corners stays under 1 px.
+- **Focus on a plane, not a point:** the focus distance is measured along
+  the camera's axis.
+- **Judge focus at full size.** Render `vos still … --width 1920 --height
+  1080` and crop at 100%: a 1280×720 still or a contact sheet hides 10 px
+  of blur.
+
 ## Design principles
 
 - **Visual impact**: bloom for glow (even 0.3–0.5 elevates); emissive

package/skills/vos-port/references/elements-and-context.md

@@ -1,4 +1,4 @@
-<!-- Generated by scripts/build-element-reference.mjs from @vosjs/core 0.25.3, @vosjs/elements 0.9.0, @vosjs/timeline 0.4.1 and @vosjs/shared 0.5.0. Do not edit by hand: re-run the script. -->
+<!-- Generated by scripts/build-element-reference.mjs from @vosjs/core 0.25.5, @vosjs/elements 0.9.0, @vosjs/timeline 0.4.1 and @vosjs/shared 0.6.0. Do not edit by hand: re-run the script. -->
 
 # Elements, the context and eases: the declarations
 
@@ -353,6 +353,12 @@ unit boundaries, the line aligned by `font.align`). So a per-letter tween
 moves each letter FROM its laid-out place (`from`, or `s.y + dy`), never
 to a position computed by hand.
 
+A split element's `props` are the WORD (`@vosjs/elements` 0.10+): `x`/`y`
+offset every unit, `opacity` multiplies each unit's own, `scale` and
+`rotation` scale and turn the word about its centre, so letters rise
+through `el.segments` while `el.props` drifts or fades the word, two
+tweens, never one per letter. Before 0.10 they moved the FIRST letter only.
+
 Writing any of `content`, `fontSize`, `fontFamily`, `fontWeight`, `fontStyle`, `letterSpacing`, `color`, `strokeColor`, `strokeWidth` on a text element's `props`
 re-rasters it (the runtime's own list), so a scramble or a counter writes
 `el.props.content` in `onFrame` and a colour change sets `el.props.color`.