Reference this skill whenever generating, treating, or reviewing a beat in the Remotion render pipeline -- documentary explainer motion graphics style (Vox/Johnny Harris genre), covering photo treatment, accent-color discipline, camera/transitions, annotation, typography, maps, and pacing. Use when building or QA-ing any beat's visual output.
Install
npx skillscat add lumeierecollection-blip/youtube/vox-style-treatment Install via the SkillsCat registry.
Vox-Style Documentary Explainer -- Reference Skill
Sourced from real research on the Vox/Johnny Harris documentary-explainer
genre -- not invented. This is the standing acceptance bar for every beat
this pipeline renders, referenced automatically, not just during one-off
fix passes.
Photo/asset treatment (per-asset, never the canvas background)
- Soft vignette pulling the eye toward center
- Halftone dots -- a genuine Vox signature, not incidental (documented
as deliberate: "1950s print dots on brand-new graphics") -- distinct
from generic grain, apply as its own texture pass - Grain (film/paper), light-leak, slight chromatic aberration at edges
-- "breaking the digital feel" - Desaturated editorial base grade (cream/navy/gray family)
- Drop shadow, grounding the cutout
- SAME grade recipe applied to every sourced photo regardless of
original source/lighting -- this is what makes many different stock
photos read as one continuous piece instead of a mismatched slideshow - Implemented via real, tested libraries (postprocessing /
@remotion/three / a verified-license LUT) -- never freehand shader math
Fallback: cutout quality fails the rembg gate
- SCOPED, does not change the general rule: normal cutout-composited
beats and typography-only beats stay exactly as established (flat
white/black canvas, no background gradient) - When a specific photo fails cutout-quality scoring, use it full-bleed
for that beat instead of discarding it -- same per-photo grade
(vignette/halftone/grain/desaturation) as any other photo - Add a gradient scrim WITHIN that full-bleed photo only, for text
legibility where the caption sits -- this lives on the photo layer,
not the canvas - Typography sits in the scrim zone, same type rules as elsewhere
Accent color system
- Exactly ONE bold accent color per channel, reserved for the single
thing that matters in a given scene - Annotation marks (circles/arrows/callouts) use that SAME accent --
annotation and palette are one system, not separate choices
Camera + transitions
- Ken Burns push/pull on photos
- Blur-ramp transition option: position move + Gaussian blur keyframed
up then back down, for a "flashback"-style cut between photo beats - Transition-specific SFX paired to what's actually happening (not a
generic whoosh on every cut)
Annotation/callout elements
- Hand-drawn-style circles/arrows/connector lines highlighting a
specific detail within a photo, stroke-drawn on (not fade-in) - Always in the single accent color, never competing with other colors
Typography
- Tight kerning, strong horizontal rhythm on titles
- Text elements slide in, not hard-cut
- Chapter/section openers styled like book sections, not a generic
YouTube intro card
Maps (where relevant)
- Route lines draw on (stroke animation), not appear instantly
- Labels animate in per-location, camera moves between points
- Same accent-color system as everything else
Structure/pacing (may require script-stage changes, not just render)
- "Visual evidence, then context" -- cold open on a visual anchor
BEFORE explaining it, rather than setup-then-payoff - Verify whether current script generation (Stage C) already produces
this ordering or needs a prompt change to prompts/write-script.md
Known tension -- decided, not open
SFX-on-every-beat sits against this genre's convention of reserving
distinctive sounds for specific moments. Resolution: every beat gets SOME
audible cue, but distinctive sounds (whooshes, projector clicks, impacts)
stay reserved for moments that actually earn them.