Skip to content

feat(registry): 11 caption style blocks#679

Open
miguel-heygen wants to merge 2 commits intomainfrom
feat/caption-styles-catalog
Open

feat(registry): 11 caption style blocks#679
miguel-heygen wants to merge 2 commits intomainfrom
feat/caption-styles-catalog

Conversation

@miguel-heygen
Copy link
Copy Markdown
Collaborator

@miguel-heygen miguel-heygen commented May 8, 2026

Summary

11 production-quality caption/subtitle style blocks for the HyperFrames registry, covering viral short-form to professional lower-thirds.

Styles

# Block Style Preview
1 cap-boxed Netflix/YouTube dark box subtitle video
2 cap-outline Stroke-to-fill reveal (Bebas Neue) video
3 cap-karaoke-sweep CapCut-style progressive white glow video
4 cap-hormozi Hormozi word pop, gold highlight video
5 cap-pill Instagram Reels gradient pills video
6 cap-marker Yellow marker sweep (light bg) video
7 cap-clean Ali Abdaal smooth fade video
8 cap-glitch RGB chromatic aberration video
9 cap-neon Layered pink neon glow video
10 cap-bounce Elastic word entrance video
11 cap-emoji Hormozi + emoji reactions video

Technical details

  • Word-level karaoke sync from transcript timestamps
  • Hard kill at every group.end
  • Deterministic GSAP timelines, no Math.random()
  • 96-128px fonts with text-stroke/shadow for readability over any background
  • Font size auto-fitting for long phrases to prevent overflow
  • Standalone 1920x1080 compositions (9s each)

Test plan

  • hyperframes validate passes with 0 errors on all 11
  • All 11 rendered to MP4 and uploaded to S3
  • Review video previews linked above

11 production-quality caption/subtitle styles for the HyperFrames registry,
covering the full spectrum from viral short-form to professional lower-thirds.

Each style was validated against a custom attention scorer (adapted from
c2v-bench) measuring readability, confidence, and contrast. The quality bar
was set at 75+ readability and 0.95+ confidence — 7 of 11 exceed this,
with the remaining 3 (neon, glitch, bounce) scoring 65-70 due to their
intentional dim-inactive/bright-active contrast design.

Styles:
- cap-hormozi: Montserrat 900, word pop, gold highlight (readability: 77.5)
- cap-karaoke-sweep: full phrase, progressive white glow (78.3)
- cap-pill: rounded gradient pills per word, red active (75.9)
- cap-neon: layered pink neon glow on active word (69.1)
- cap-boxed: Netflix-style dark box subtitle (81.7 — highest)
- cap-outline: stroke-to-fill reveal, Bebas Neue (79.7)
- cap-bounce: elastic word entrance, cyan highlight (64.6)
- cap-glitch: RGB chromatic aberration entrance (69.9)
- cap-marker: yellow marker sweep on light background (75.6)
- cap-clean: Ali Abdaal smooth fade, zero decoration (75.5)
- cap-emoji: Hormozi pop with contextual emoji reactions

All styles use:
- fitTextFontSize (pretext) for overflow prevention
- Hard kill (tl.set opacity:0, visibility:hidden) at every group end
- Word-level karaoke sync from transcript timestamps
- 96-128px font minimum with text-stroke/shadow for readability
- Deterministic GSAP timelines (no Math.random)

Video previews: s3://heygen-public/hyperframes/caption-blocks/
@miguel-heygen miguel-heygen changed the title feat(registry): 11 caption style blocks — attention-scored, pretext-safe feat(registry): 11 caption style blocks May 8, 2026
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