Skip to content

release: promote develop to main (v0.10.2)#903

Merged
chronoai-shining merged 8 commits into
mainfrom
develop
Jun 5, 2026
Merged

release: promote develop to main (v0.10.2)#903
chronoai-shining merged 8 commits into
mainfrom
develop

Conversation

@chronoai-shining

Copy link
Copy Markdown
Collaborator

Related to #901

Promotes developmain to cut v0.10.2.

Carries:

On merge, the changeset-release workflow will open release/v0.10.2 → main (review + squash-merge), then tag, publish the GitHub Release, and open the sync/post-release-v0.10.2 → develop PR (merge with a merge commit, never squash).

🤖 Generated with Claude Code

chronoai-shining and others added 8 commits June 5, 2026 14:55
chore: sync main → develop after v0.10.1
Move ornn-intro.mp4 + ornn-intro-poster.jpg from public/ into src/assets/
and reference them via Vite asset imports in HeroVideo (tests follow the
imported URLs instead of hardcoded /public paths).

Why: nginx serves images with 'expires 1y + immutable', so replacing the
poster bytes at a stable /public URL leaves returning visitors on the
stale cached file. Content-hashed URLs make every future asset swap
self-cache-busting; this also unlocks long-lived caching for the video
itself (today it falls through to the no-store SPA fallback location and
re-downloads on every visit).

Pure refactor: same bytes, same rendered markup semantics.
The hero video previously fell through to the SPA fallback location,
whose 'no-cache, no-store' header forced every landing visit to
re-download the ~20MB file. Now that video assets ship under Vite
content-hashed URLs (previous commit), mp4 can join the 1y-immutable
static-asset location: a swap changes the URL, so a cached entry can
never go stale.
Replace the hero video + poster with a clean re-render from the
ornn-marketing Remotion source (new OrnnIntroLandClean composition):

- No caption layer: chapter cards, subtitles, the closing CTA block and
  the ORNN corner wordmark are gone — the hero is pure animation. The
  FootageFix cover for the baked-in source-footage typo stays. The page
  DOM CTA is now the only text on the hero.
- 60 fps (was 30): scene-split parallel minterpolate (mci/aobmc/bidir)
  from the 24 fps master — splitting at the 8 hard cuts avoids cross-cut
  morph frames. Same 2x fast-intro pacing as before (~82s loop).
- True 16:9 1920x1080 frame (was 2592x1080): the old 2.4:1 canvas baked
  blurred letterbox bars into the pixels, which read as dead side
  gutters on wide viewports. object-cover now fills every viewport
  ratio with real content pixels.
- H.264 High@L4.2 (1080p60 ceiling), yuv420p, two-pass 2.8 Mbps, ~29 MB,
  faststart, no audio — decodable in hardware on phones/laptops (the
  constraint that reverted the 4K cut in #870).

Comments in HeroVideo updated: the caption-safety rationale (AC4 of
#840) no longer applies; CTA/scrim placement is unchanged.
…an-60fps

feat(web): caption-free 60fps true-16:9 landing hero video
Curated release-notes-20260607.md for v0.10.2 (caption-free 60fps
landing hero video). Dated 20260607 because 20260605/20260606 are
taken by v0.10.0/v0.10.1 — the release workflow uses the most-recent
dated file as the GitHub Release body.

Includes an empty changeset to satisfy check-changeset.
@chronoai-shining chronoai-shining merged commit 5f4fdac into main Jun 5, 2026
20 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant