← GPT-AI-VIDEO
RESEARCH · 3.8 KB

research/keyframe-self-review-v1.md

Workspace snapshot · 09/04 14:52

AI-native Vertical B-roll — Keyframe Self Review v1

Date: 2026-08-30
Brief: 技術・経営思想の冒頭フック。1つの試行が複数経路へ分かれ、検証を経て1つの方向へ収束する。
Privacy: Public-safe generated plates. No first-party or third-party source assets used.

Shared decisions

  • Audience action: stop scrolling, read the later Remotion title, then continue into the editorial short.
  • Dominant emotion: curiosity followed by momentum and resolution.
  • Visual protagonist: a single controlled path or physical sequence.
  • Brand tension: disciplined physical craft versus abstract AI cliché.
  • Motion affordance: sequential illumination, unfolding, rotation, parallax, and camera passage.
  • Typography: none baked in; exact Japanese type is reserved for Remotion.

Candidate A — tech-paths-hook-v1.png

  • Intended role: hook / branching possibility.
  • Source assets: none; generated clean plate.
  • Mode: new generation.
  • Strongest feature: one cyan path creates immediate hierarchy and credible sequential-light motion.
  • Remaining risk: the visual metaphor is not yet uniquely tied to Takumi/PENGIN and could be read as premium technology stock.

Score: 22/24

DimensionScore
Source specificity1
One-second hierarchy2
Emotional clarity2
Light quality2
Composition tension2
Palette discipline2
Subject integrity2
Typography readiness2
Motion affordance2
Mobile reading2
Brand integrity1
Anti-template2

Blockers: none.
Decision: shortlist for independent review; do not send to video yet.

Candidate B — prototype-sequence-development-v1.png

  • Intended role: development / iteration becoming tangible.
  • Source assets: none; generated clean plate.
  • Mode: new generation.
  • Strongest feature: physical paper construction avoids generic hologram imagery and offers sequential unfolding.
  • Remaining risk: no single protagonist is obvious at phone size; a star-like artifact in the lower-right corner reads as a possible watermark or decorative generator mark.

Score: 18/24

DimensionScore
Source specificity1
One-second hierarchy1
Emotional clarity1
Light quality2
Composition tension2
Palette discipline2
Subject integrity1
Typography readiness2
Motion affordance2
Mobile reading1
Brand integrity1
Anti-template2

Blockers: possible watermark/decorative star; weak protagonist.
Decision: reject. If this concept is revisited, regenerate only the hierarchy/artifact defect rather than changing the whole style.

Candidate C — convergence-aperture-payoff-v1.png

  • Intended role: payoff / many paths resolving into one.
  • Source assets: none; generated clean plate.
  • Mode: new generation.
  • Strongest feature: the cobalt aperture and strong foreground ring create a credible camera move and loopable convergence action.
  • Remaining risk: architectural sculpture is visually strong but still insufficiently specific to the customer story.

Score: 22/24

DimensionScore
Source specificity1
One-second hierarchy2
Emotional clarity2
Light quality2
Composition tension2
Palette discipline2
Subject integrity2
Typography readiness2
Motion affordance2
Mobile reading2
Brand integrity1
Anti-template2

Blockers: none.
Decision: shortlist for independent review; do not send to video yet.

Current outcome

  • 2/3 candidates pass the self-score threshold of 20/24.
  • 1/3 is rejected with a blocker.
  • No candidate is promoted to paid image-to-video until independent Claude review is returned and factual review claims are checked against the files.
  • The visual system still needs stronger source specificity before being called a validated brand asset.