Planning

Your mantra: music is the spine; a template is a head start, not a cage; typography is the floor and assets are an optional ingredient on the same beat grid.

You own WHAT, not HOW. You name the template / primitives, the content, the brand, the anchor seconds, and the intent. The frame-worker (Step 4) decides HOW — micro-timing, realization, intra-frame cuts. Never write millisecond tweens into the storyboard.

Inputs

• The Step-2 skeleton already in STORYBOARD.md — frames with span_sec + pacing + mood + feel.
• audiomap.json — timing truth; read the real anchor seconds inside each frame’s span.
• template-catalog.md — the template selection menu.
• motion-primitive-catalog.md — the free-compose menu (L0 recipes).
• montage.md — asset treatments (only if the user supplied images/videos).
• User brief / supplied copy — topic, mood, exact words to keep.

Step A — pick the brand spine (one preset, unmodified)

The whole video shares one type family + palette. Pick one ready-made preset from ../hyperframes-creative/frame-presets/ using the preset table in ../hyperframes-creative/references/design-spec.md — choose by the track’s mood + the brief, and only its fonts + colors matter (templates own composition + motion; the preset only sets the look). Copy it in unmodified:

Then fill the storyboard frontmatter style from it: the font from its typography: and a ≤4–6 swatch palette from its colors:. Quote the hex / family verbatim — never invent or round. Every group’s palette params draw from this one palette; that unity is what makes different templates read as one piece.

Step B — per frame, decide its groups

A frame is usually one group (one template or one free composition spanning the frame). Subdivide into 2+ groups only when a single treatment can’t cover the frame — e.g. a busy opener plus a closing lockup. When you split, cut at a real audiomap anchor inside the frame’s span (a key_moment / phrase edge / onset-cluster gap), never inside a rolls[] run, and keep every group ≥ ~1 bar. Density does not force more groups — a dense frame is usually ONE group whose template absorbs the density internally (a meta-template like poster-tile-mosaic). Group count tracks distinct treatments, not beats.

Step C — per group, pick a treatment (exactly one of three)

A. Match a template

Read template-catalog.md. Match the group’s feel + mood + pacing to a template’s Reach for it when; take the closest fit. Then bind it:

• Fill params (keys from the catalog entry) — your copy into text slots, palette from the brand spine, duration = the group’s span length.
• Fill role_bindings with this group’s real anchor seconds read from audiomap.json over its span (not example times).
• If the template’s natural stop and the group’s span end disagree, snap to the nearest anchor.

B. Free-compose (no template fits)

Write a free_design — one visual thesis from motion-primitive-catalog.md (a dominant system + the named L0 primitives + a density topology) + anchors (the real beat / onset seconds the moves ride). Free-compose is a first-class choice, written as carefully as a matched group — never a failure.

C. Asset treatment (only when the user supplied assets and they fit)

Make it an asset group (montage.md). Obey pacing: on a beat_cut frame use beat_cut (one clip per anchor) or bg_under_text; on a phrase_flow frame use ken_burns or a slow crossfade — never per-onset hard cuts. Assets are additive: if none fits a group, fall back to template / free (typography is the floor — a complete video needs no assets).

Copy (you own the words)

• Keep exact user words; else invent with taste, on the brief’s mood.
• Message vs texture: a readable word holds ≥1 beat (headline 3–8, sentence 4–10), stable + focal; a word held <1 beat is texture (strobe / grid / ticks). Never force a message onto a sub-beat — demote it to texture.
• Place copy into the template’s text params, onto a free group’s anchors, or as an asset group’s overlay_copy. Declare the anchor + accumulate / stagger intent; leave micro-timing to the worker.
• A closing logo / CTA lands on the final hit / hard stop and holds through trailing silence.

Transitions (you do not emit them)

Everything is a 0ms hard cut for now. frame → frame is owned by the assembler (back-to-back files); adjacent span_sec already imply the cut. group → group inside a frame is owned by the worker on its frame timeline; you only set each group’s span_sec.

Write + validate

Complete STORYBOARD.md (storyboard-format.md), then run node scripts/validate-plan.mjs and fix every ✗. Present the frame-by-frame summary in chat and iterate until approved (Step 3 in SKILL.md says how).

Self-check

• frame.md is a verbatim copy of one preset; frontmatter style.font / style.palette are drawn from it (exact values).
• Every frame became ≥1 group; groups tile the frame span in order; no group < ~1 bar; no group boundary inside a rolls[] run.
• Each group is exactly one of template / free_design / asset.
• Template params keys match the catalog entry; role_bindings / anchors use real audiomap seconds.
• Asset treatments obey pacing (no beat_cut on a phrase_flow frame).
• Every group’s palette draws from the one brand palette.
• duration_s == audiomap.audio.duration_sec; validate-plan.mjs passes.