reelkit-cli 0.12.4 → 0.12.6

package/CHANGELOG.md

@@ -3,6 +3,16 @@
 What changed in each published version of `reelkit-cli`, newest first. A push to `main` publishes the version in
 `package.json` when it is not on npm yet, and it must have an entry here.
 
+## 0.12.6 - 2026-10-10
+
+- `reelkit template push` leaves out `.remotion` (the browser an older project rendered with) and `.cache`, and when a project is still over the limit it names the largest folders.
+
+## 0.12.5 - 2026-10-09
+
+- The skill builds a film from component files instead of one long `Video.tsx`: every drawn part that could appear in another film is its own file in `src/`, so it can be reused and shows as a part of a template.
+- A template lists the components a composition declares for itself, and ties a component to a scene when its name is the scene's id.
+- A template carries the spoken words of each scene, for a subtitles row on its timeline, and the sound effects the project uses, so a sound on the timeline copies the command that pulls it.
+
 ## 0.12.4 - 2026-10-09
 
 - A template's timeline shows the music and its beats to the end of the film when the track repeats, not only for the track's own length.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "reelkit-cli",
-  "version": "0.12.4",
+  "version": "0.12.6",
   "description": "CLI and Claude skill for making short-form video with a shared asset library.",
   "license": "MIT",
   "author": "Daniel Livshin",

package/skill/SKILL.md

@@ -15,7 +15,7 @@ For image or video cutouts, editable Blender product shots, custom 3D materials,
 
 To implement a chosen style, read its row in `reference/style-recipes.md` for action, timing, sound and proof. `reference/style-components.md` maps the shared components and built-in kit to shot roles; inspect the selected source before using its props.
 
-This skill targets `reelkit-cli` 0.12.4 and its HyperFrames kit. Check the installed version with `reelkit --version`; in a source checkout, invoke its `bin/reelkit.mjs` with Node. Use the matching CLI before authoring with `reelkit/frame`.
+This skill targets `reelkit-cli` 0.12.6 and its HyperFrames kit. Check the installed version with `reelkit --version`; in a source checkout, invoke its `bin/reelkit.mjs` with Node. Use the matching CLI before authoring with `reelkit/frame`.
 
 ## What decides whether the film is accepted
 
@@ -133,6 +133,7 @@ Read `reference/hyperframes-composition.md` for the seekable HyperFrames timing
 - For a brand reveal, a transformation, or a shot with real depth, read `reference/brand-motion.md` and `reference/three-d.md`. Choose the move by what it explains, preserve the supplied identity, and share its event frames with the sound cues. Keep reading-heavy shots in 2D; a requested brand film or showreel may use several purposeful 3D moments.
 - For what sits behind the whole film (a video, a picture that changes, or an animated ground), read `reference/backgrounds.md`; pick one ground and keep it.
 - Before writing a new component, read `reference/component-authoring.md`.
+- **Build the film from component files, not one long `Video.tsx`.** Every drawn part that could appear in another film (a text reveal, a card, a character, a chart, a cursor, a shape that morphs, a field of particles) is its own file in `src/`, driven by props, with the comment that describes it. `Video.tsx` keeps only what belongs to this film: the palette, the words, the frame numbers, where each part sits in each scene, and the sound. A helper declared inside `Video.tsx` that returns JSX and holds none of this film's words is a component that was not split out; move it. A film of several scenes should end with several component files, and they are what the user reuses in the next video and what a template shows as its parts.
 - Write `src/Video.tsx` and any component files beside it. `Video.tsx` may import only `react`, `reelkit/frame`, `reelkit/kit` and sibling components (`./Name`).
 - Run `reelkit lint --no-video` together with `reelkit check`. It reads the source: right-to-left text with no direction, vowel points in on-screen text, a sequence that starts after its parent ends, and a shot written on a range the film never draws.
 - `manifest.json` is the exact object passed to `Video` as the `manifest` prop. Media is referenced as `urls[path]`, where `path` is the file's path in the project, such as `urls[scene.voiceoverKey]` or `urls["assets/lib/<id>/clip.mp3"]`.

package/skill/reference/component-authoring.md

@@ -10,6 +10,25 @@ Run `reelkit assets search "<description>" --kind component`. If a library compo
 
 Use `reference/style-components.md` to choose the role and `reference/style-recipes.md` for the shot's motion and proof. A catalog name does not mean a built-in kit export or a currently published asset; search, pull and inspect the actual source first.
 
+## Split the film into parts
+Searching first is about not writing a part twice. It is not a reason to write everything inside `Video.tsx`: a film whose whole picture lives in one file leaves nothing to reuse, shares nothing, and shows as one block when it is saved as a template.
+
+Before writing `Video.tsx`, list the parts the film draws and decide where each one lives:
+
+| It is | It goes | Example |
+|---|---|---|
+| A drawn thing with no words of its own | Its own file in `src/`, all values as props | `MaskRise` (a line that rises from behind a mask), `Blob`, `EyedCharacter`, `ChartCard`, `TabChip`, `DrawnCheck` |
+| A way of placing or timing other things | Its own file, taking `children` | `Pin` (centre children on a point), `Shake`, `Orbit` |
+| This film's scene: which parts, which words, which frames | `Video.tsx` | the `hook` shot: a `ChartCard` at frame 4, "Great month." at 53, the slam at 98 |
+| The palette, the sound list, the frame numbers | `Video.tsx`, passed down as props | the blob gets `color={c.hero}` |
+
+Rules of thumb:
+- A function in `Video.tsx` that returns JSX and holds none of this film's words is a part. Move it to a file.
+- A scene function may stay in `Video.tsx`, named after its scene id (`Hook` for `hook`, `OneNumber` for `one-number`), and should be mostly a list of parts with props. If it is long because it draws shapes itself, the shapes are parts.
+- One part needs another (a monster has eyes): pass it in as `children` or a prop, since a component file imports no other component file.
+- The same thing lives through several scenes (one ball that changes shape): one part with the values as props (`x`, `y`, `radius`, `morph`), and `Video.tsx` owns the keyframes.
+- Do not split for its own sake: a two-line wrapper used once is not a part.
+
 ## Rules for a component file
 - One component per file. File `NumberBadge.tsx` exports `export const NumberBadge: React.FC<NumberBadgeProps>`.
 - Imports only from `react`, `reelkit/frame` and `reelkit/kit`. A component never imports another component file.

package/src/commands/template.ts

@@ -21,7 +21,8 @@ const sh = promisify(execFile);
 // What is sent: the project folder as one archive (the plan, the composition, and every file under assets/, including the user's own
 // pictures, footage and the recorded voice), the rendered video when there is one, and a small picture of each scene. What is left on
 // this machine: out/ (renders and previews), refs/ (reference videos), caches and node_modules.
-export const TEMPLATE_SKIP = ["out", "refs", ".reelkit", "node_modules", ".git", ".DS_Store"];
+// .remotion is where a project made before 0.10 kept the browser it rendered with: hundreds of megabytes that are not the project.
+export const TEMPLATE_SKIP = ["out", "refs", ".reelkit", ".remotion", ".cache", "node_modules", ".git", ".DS_Store"];
 const VIDEO = "out/video.mp4";
 // Where a project remembers the template it was cloned from, and the last one it was pushed as. Never part of a bundle.
 const LINK = ".reelkit/template.json";
@@ -56,7 +57,7 @@ export function filmOf(project: Project): TemplateFilm | undefined {
       return {
         id: cut(m.id, 80), start: Math.max(0, Math.round(m.startFrame)), frames: Math.max(0, Math.round(m.durationFrames)),
         ...(p?.treatment ? { treatment: cut(p.treatment, 60) } : {}), ...(narration ? { narration: cut(narration, 4000) } : {}),
-        ...(text.length ? { text } : {}), ...(m.words.length ? { words: m.words.slice(0, 600).map((w) => Math.max(0, Math.round(w.startSec * 1000) / 1000)) } : {}),
+        ...(text.length ? { text } : {}), ...(m.words.length ? { words: m.words.slice(0, 600).map((w) => Math.max(0, Math.round(w.startSec * 1000) / 1000)), spoken: m.words.slice(0, 600).map((w) => cut(String(w.word ?? ""), 80)) } : {}),
         ...(p?.notes?.trim() ? { notes: cut(p.notes.trim(), 2000) } : {}),
         ...(used.some((c) => c.scenes.includes(m.id)) ? { components: used.filter((c) => c.scenes.includes(m.id)).map((c) => cut(c.name, 80)).slice(0, 60) } : {}),
       };
@@ -66,9 +67,18 @@ export function filmOf(project: Project): TemplateFilm | undefined {
     voice: plan.voice !== "none",
     ...(manifest.captions ? { captions: manifest.captions } : {}),
     ...cuesOf(project, manifest, music?.key),
+    ...soundsOf(project),
   };
 }
 
+// The sounds the project pulled from the library, by id: what a timeline can offer to copy when it cannot say which sound a hit is.
+function soundsOf(project: Project): Pick<TemplateFilm, "sounds"> {
+  const sounds = Object.entries(project.readJsonOr<Record<string, { kind?: string; title?: string }>>(FILES.library, {}))
+    .filter(([, p]) => p.kind === "sfx" || p.kind === "music")
+    .map(([id, p]) => ({ id: id.slice(0, 128), title: (p.title ?? id).slice(0, 200), kind: p.kind! })).slice(0, 300);
+  return sounds.length ? { sounds } : {};
+}
+
 // The music as it plays in the film. A track shorter than the film is repeated by the composition, and the manifest holds every beat
 // that falls inside the film (the repeats too), so those are the beats shown; the track then runs to the film's end. Without them the
 // track's own beats and length are used.
@@ -147,6 +157,13 @@ export function filesOf(dir: string): { files: TemplateFile[]; sources: Uint8Arr
   return { files, sources: sources && sources.length <= MAX_TEMPLATE_SOURCES_BYTES ? sources : undefined };
 }
 
+// The three largest things at the top of the project, for the message that says a project is too large to push.
+function largest(dir: string): string {
+  const size = (path: string): number => { try { const st = statSync(path); return st.isDirectory() ? readdirSync(path).reduce((sum, name) => sum + size(join(path, name)), 0) : st.size; } catch { return 0; } };
+  const top = readdirSync(dir).filter((name) => !TEMPLATE_SKIP.includes(name)).map((name) => ({ name, bytes: size(join(dir, name)) })).sort((a, b) => b.bytes - a.bytes).slice(0, 3).filter((x) => x.bytes > 1_000_000);
+  return top.length ? `The largest parts: ${top.map((x) => `${x.name} (${mb(x.bytes)})`).join(", ")}. ` : "";
+}
+
 const readLink = (project: Project): { id?: string; from?: string } => { try { return project.readJsonOr<{ id?: string; from?: string }>(LINK, {}); } catch { return {}; } };
 
 export async function templatePush(ctx: Ctx, opts: { name?: string }): Promise<Result> {
@@ -162,7 +179,7 @@ export async function templatePush(ctx: Ctx, opts: { name?: string }): Promise<R
     ctx.log("Packing the project.");
     await sh("tar", ["-czf", bundle, ...TEMPLATE_SKIP.flatMap((x) => ["--exclude", `./${x}`]), "-C", project.dir, "."], { maxBuffer: 1 << 24 });
     const bundleBytes = statSync(bundle).size;
-    if (bundleBytes > MAX_UPLOAD_BYTES) return { ok: false, summary: `The project packs to ${mb(bundleBytes)}, over the ${mb(MAX_UPLOAD_BYTES)} a template may be. Remove large files under assets/ that the video does not use and push again.` };
+    if (bundleBytes > MAX_UPLOAD_BYTES) return { ok: false, summary: `The project packs to ${mb(bundleBytes)}, over the ${mb(MAX_UPLOAD_BYTES)} a template may be. ${largest(project.dir)}Move out what the video does not use and push again.` };
     const video = project.exists(VIDEO) ? project.path(VIDEO) : undefined;
     const videoBytes = video ? statSync(video).size : 0;
     if (videoBytes > MAX_UPLOAD_BYTES) return { ok: false, summary: `The rendered video is ${mb(videoBytes)}, over the ${mb(MAX_UPLOAD_BYTES)} limit. Export a smaller one with \`reelkit export\` or push without a render.` };

package/src/contract/index.ts

@@ -194,6 +194,8 @@ export const TemplateFilmSchema = z.object({
   scenes: z.array(z.object({
     id: text(80), start: frames, frames, treatment: text(60).optional(), narration: text(4000).optional(),
     text: z.array(text(400)).max(20).optional(), words: z.array(z.number().min(0)).max(600).optional(),
+    // The spoken words themselves, one for each entry of `words`, so the captions can be laid out in time.
+    spoken: z.array(text(80)).max(600).optional(),
     // The plan's visual direction for the scene, and the components the composition times to it (read from its source).
     notes: text(2000).optional(), components: z.array(text(80)).max(60).optional(),
   })).min(1).max(200),
@@ -209,6 +211,8 @@ export const TemplateFilmSchema = z.object({
   // The sound effects by name, when the render recorded them: each one's title, when it starts and how long it plays, in seconds.
   // `id` is the sound's id in the library, when it was pulled from there.
   cues: z.array(z.object({ at: z.number().min(0), seconds: z.number().min(0), title: text(200), kind: text(24).optional(), id: text(128).optional() })).max(1500).optional(),
+  // The sounds the project pulled from the library (effects and music), whether or not the render recorded when each one plays.
+  sounds: z.array(z.object({ id: text(128), title: text(200), kind: text(24) })).max(300).optional(),
   // Every component the composition uses: one written in the project, one pulled from the library, or one of the kit's. `uses` is how
   // many times it appears, and `about` is the comment above a project component.
   components: z.array(z.object({ name: text(80), from: z.enum(["project", "library", "kit"]), uses: z.number().int().min(1), about: text(400).optional() })).max(300).optional(),

package/src/project/scene-components.ts

@@ -103,7 +103,9 @@ export function componentsOf(source: string, opts: { sceneIds: string[]; pulled?
     for (const [name, body] of local) if (body === node) { inside = name; here = new Set(); }
     const opening = ts.isJsxSelfClosingElement(node) ? node : ts.isJsxElement(node) ? node.openingElement : undefined;
     if (opening) {
-      const own = idsIn(opening.attributes);
+      // A spread of props ({...fx}) is a bundle, often of every scene's timings, so it names no scene: only the props written out do.
+      const own = new Set<string>();
+      for (const attr of opening.attributes.properties) if (!ts.isJsxSpreadAttribute(attr)) for (const id of idsIn(attr)) own.add(id);
       here = own.size ? own : here;
       const tag = opening.tagName.getText(sf);
       const root = tag.split(".")[0]!;
@@ -154,10 +156,40 @@ export function componentsOf(source: string, opts: { sceneIds: string[]; pulled?
       for (const use of list) for (const id of use.scenes.size ? use.scenes : (scenesOfLocal.get(use.owner) ?? [])) if (!mine.has(id)) { mine.add(id); changed = true; }
     }
   }
+  // A component of the file's own that is named after a scene ("Hate" for the scene "hate", "Good" for "too-good") is that scene's,
+  // however it is timed inside.
+  // Its name decides alone: whatever else its props mention, "Hate" is the scene "hate".
+  const plain = (text: string) => text.toLowerCase().replace(/[^a-z0-9]/g, "");
+  const byName = new Set<string>();
+  for (const name of scenesOfLocal.keys()) {
+    const n = plain(name);
+    const hits = [...known].filter((id) => { const i = plain(id); return n.length >= 3 && (i === n || (n.length >= 4 && (i.endsWith(n) || i.startsWith(n))) || (i.length >= 4 && (n.endsWith(i) || n.startsWith(i)))); });
+    if (hits.length) { scenesOfLocal.set(name, new Set(hits)); byName.add(name); }
+  }
+  // With scenes named this way, where a component sits says more than what its props mention (inside "Hate", everything is timed
+  // from the same handful of numbers): it belongs to the scenes of the component it is written in, and only a component written
+  // straight in the composition is read from its own props. Worked out again from the start on that rule.
+  const sceneFor = (use: Use): Iterable<string> => {
+    const around = byName.size ? scenesOfLocal.get(use.owner) : undefined;
+    return around?.size ? around : use.scenes.size ? use.scenes : (scenesOfLocal.get(use.owner) ?? []);
+  };
+  if (byName.size) {
+    for (const name of scenesOfLocal.keys()) if (!byName.has(name)) scenesOfLocal.set(name, new Set());
+    for (let changed = true, rounds = 0; changed && rounds < 12; rounds++) {
+      changed = false;
+      for (const [name, list] of localUses) {
+        if (byName.has(name)) continue;
+        const mine = scenesOfLocal.get(name)!;
+        for (const use of list) for (const id of sceneFor(use)) if (!mine.has(id)) { mine.add(id); changed = true; }
+      }
+    }
+  }
   const used = new Map<string, { uses: number; scenes: Set<string> }>();
+  // The file's own components are listed too: they are what most compositions are made of.
+  for (const name of local.keys()) { const list = localUses.get(name) ?? []; if (list.length) { used.set(name, { uses: list.length, scenes: scenesOfLocal.get(name)! }); origin.set(name, "project"); } }
   for (const [tag, list] of uses) {
     const scenes = new Set<string>();
-    for (const use of list) for (const id of use.scenes.size ? use.scenes : (scenesOfLocal.get(use.owner) ?? [])) scenes.add(id);
+    for (const use of list) for (const id of sceneFor(use)) scenes.add(id);
     used.set(tag, { uses: list.length, scenes });
   }