fix(runtime): respect hidden ancestor clips in Studio preview (#1387)

[manupareekk]

Summary

Fixes #1387 — Studio preview showed all caption/telop content stacked simultaneously when GSAP tweens target children of timed clips (the standard caption pattern).

Two changes in packages/core/src/runtime/init.ts:

• Skip Studio timing stamps on GSAP targets / [id] elements that live inside an authored timed clip ancestor (prevents full-duration pseudo-clips)
• Respect hidden ancestor clips when applying inline visibility — descendants no longer get visibility: visible when a parent timed clip is hidden

Test plan

• bun run --filter @hyperframes/core test src/runtime/init.test.ts (23 tests)
• Regression test for issue Studio preview: descendants of hidden timed clips become visible when they are GSAP tween targets (clips "stack" on top of each other) #1387 minimal caption repro
• Regression test for multi-panel beat stack (niemmo launch-video pattern)
• Existing nested-composition visibility tests still pass
• Manual: hyperframes preview on a composition with caption children — only one caption visible at a time in Studio

[miguel-heygen]

Strengths

• findTimedClipAncestor (init.ts:400–411) correctly identifies the stamping root cause: Studio was iterating GSAP tween targets and stamping any element lacking data-start with a full-composition-duration pseudo-clip, regardless of position in the authored hierarchy. The guard at init.ts:1070 and init.ts:1090 eliminates that injection entirely for tween targets inside authored timed clips.
• Extracting the old inline visibility-window logic into isTimedElementVisibleAt is a clean refactor with zero behavior risk — it's a direct extraction and re-use.
• Three regression tests cover the exact failure modes: per-caption stacking (#1387), no-stamp-in-Studio, and the niemmo multi-panel beat pattern. Fixtures are minimal and readable.

Important

Fix 2 (ancestor visibility walk) changes render-mode behavior, not just Studio preview. The seek handler that now walks ancestor data-start elements runs in both Studio and render contexts. If any existing composition intentionally has a child element with a longer time window than its parent timed clip — e.g. a headline that bleeds out slightly past the parent panel's official data-duration — it would now be suppressed when the parent goes hidden. The new tests only cover the caption/panel-stack pattern. Were any existing compositions checked for this pattern before landing?

This isn't a blocker given the bug's severity, but the behavioral scope is wider than the PR description implies.

Nit

findTimedClipAncestor is called with rootComp potentially null (resolveRootCompositionElement() can return null). When null, the guard if (node === rootComp) break becomes if (node === null) break, which the while (node) loop already handles — so it's safe, but a comment acknowledging the null case would reduce reader confusion.

---

Verdict: APPROVE
Reasoning: Fix 1 eliminates the root cause (spurious Studio stamps on tween targets inside authored clips); Fix 2 enforces correct hierarchical visibility semantics as defense-in-depth. Together they're complete. The important note above is worth the author confirming, but it doesn't gate merge.

— Magi

[manupareekk]

Good catch on render scope — pushed a follow-up that gates the ancestor visibility walk to window.parent !== window, matching the stamping fix. Render mode keeps legacy per-element semantics; added a regression test for a child with a longer authored window than its parent. Also added the null rootComp comment. Fix 1 alone eliminates the root cause; Fix 2 is now Studio-only defense-in-depth.

[miguel-heygen]

Follow-up on the new commit (a3cecdb).

My "important" flag is fully resolved. The ancestor visibility walk is now gated by window.parent !== window (init.ts:1505) — it only fires in the Studio iframe, leaving render mode's per-element visibility semantics untouched. The new render-mode regression test confirms exactly the case I was asking about: a child with a wider data-duration than its parent is not suppressed in render mode.

Null-rootComp comment added — nit cleared.

withStudioIframe helper is a nice cleanup over the inline Object.defineProperty pattern in the original tests.

Approval stands.

— Magi

[miguel-heygen]

Merge activity

• Jun 12, 11:57 PM UTC: Graphite couldn't merge this PR because it failed for an unknown reason (Fast-forward merges are not supported for forked repositories. Please create a branch in the target repository in order to merge).