Aesthetic Principles

The 18 rules

1. The subject always wins

Captions exist to serve the face, not compete with it. If the caption draws the eye before the speaker, the caption has failed. Before shipping any plan.json, ask: “does the caption or the subject read first?” If caption wins, shrink, dim, or move it.

2. Occlude, don’t hover

World-class caption design puts text into the scene. Letters that pass behind a shoulder, mic, or head feel diegetic. Letters that float uniformly above the lower-third feel like PowerPoint. Use the matte pipeline — that’s the moat over every competitor.

3. Contrast is a hierarchy problem, not a brightness problem

A white box behind text is a failure of taste. Priority order:

1. mix-blend-mode: overlay or screen picking up scene luminance
2. 2–3px dark stroke + soft drop shadow
3. Narrow semi-opaque gradient bar (NOT a solid box)
4. Dim the background plate by 10–15% locally
5. (LAST RESORT) A hard white pill box — banned on cinematic directions

4. Kill the constant lower-third

Fixed-bottom captions are monotone. Caption zone shifts with shot:

• Tight close-up → upper sidebar or crown
• Mid-shot → embedded on back wall / foam / whiteboard
• Wide → classic lower-third offset to non-subject side

5. One family, two weights maximum

Hierarchy lives in weight (e.g., 500 → 800), not in font. Mixing Montserrat + script + serif in one clip is the #1 amateur tell. Ship Inter, SF Pro, Söhne, GT America, Aktiv Grotesk, or Neue Haas Grotesk with ≥5 weights and compose with weight + size.

6. Tracking tightens as size grows

Display-size (>40pt) wants negative tracking (-10 to -30 units, or -0.015em to -0.035em). Body-size (14–20pt equivalent) wants positive tracking (+5 to +15, or +0.005em to +0.015em). Apple SF Pro’s optical-size model is the reference. Submagic defaults do the opposite — they look cheap because of it.

7. Cap height ≈ 3.5–5% of frame height for body captions

• 1080p vertical (1920 tall): body 65–95px, emph/hook 130–170px
• 1080p landscape (1080 tall): body 40–55px, emph 70–100px
• Hormozi-size (~7–9% of frame) is hook size, not body size. Reserve for punchlines.

8. Never italicize for emphasis in video text

Italics are a print convention for flow inside a paragraph. On 24fps motion they read as “tilted” not “stressed”. Emphasize with weight (extrabold), color (single accent), or size (1.3–1.6×). Italics allowed only for:

• Literal quotation of written material
• Foreign word
• Thought vs spoken dichotomy

9. Color discipline: one hue + neutrals

Pick one saturated accent per video for keyword highlights. Hormozi’s yellow+green+red works for him because his content is already loud. Cinematic = single accent + white/bone/charcoal. Default palette:

• Warm white #F5EFE6 on dark
• Graphite #1A1A1A on light
• Accent chosen from scene sampling

10. Animate transform only, never the letter itself

letter-spacing, filter:blur, font-weight animations cause inline-block reflow → line-jumps. Animate translateY, scale, opacity, clip-path. This is locked from the embedded-captions debugging.

11. Minimum 0.4s per word visible

BBC reading speed is 160–180 wpm (0.33–0.38s/word). Cinematic feel wants more breathing room — 200–220ms stagger per word, hold full phrase 0.4–0.6s before exit. Entries under 150ms read as frantic.

12. Stagger is the primary expressive axis

Same phrase, different stagger, totally different feel:

Pick from content tone, not default to one value.

13. Emphasis escalates within a phrase, not between phrases

Every word bolded = no emphasis. Structure:

• 70% plain body
• 20% slight lift (color OR weight, not both)
• 8% full emphasis (bigger, brighter, held longer)
• 2% climax (biggest, held 1.5s, breath before + after)

14. Break rhythm once per 30s

Every-word-same-way = eye adapts, stops registering motion. Plant a rhythm-break every ~30s:

• Phrase enters from opposite direction
• A single word at 2× size
• A beat of pure silence with no caption
• A color shift

This is what separates Submagic-preset work from something designed.

15. Caption what adds, cut what restates

Transcribe everything. Display 70–85%. Remove:

• Filler (“um”, “like”, “you know”, “I mean”)
• Self-corrections (“I think… I mean actually…”)
• Obvious visual echoes (“as you can see here” while pointing)

Editorial judgment — no existing AI caption tool does this. It’s pure upside.

16. Segment on breath, not on duration

Chunk at natural pauses ≥ 250ms. A caption spanning a breath-break feels wrong. Whisper word-level timestamps make this trivial.

17. Safe zones per platform, always

• Use the safe boxes in /hyperframes-studio (Safe zones) for wide and vertical framings; that skill owns the values.

Bake into the layout solver. Never eyeball.

18. Letterbox/pillarbox is caption real estate

Black bars on 9:16 from 16:9 source? Those bars are the caption home. 2.35:1 cinematic frame on 16:9 export? Serif quotation in the letterbox reads as documentary.

Rhetorical judgment rules (editorial)

These are the “what to caption” decisions. No existing tool exercises them.

• Filler suppression — off/light/strong. Default: light for documentary, strong for vlog.
• Self-correction folding — display only the final version.
• Breath-group segmentation — split on silences >250ms, never mid-phrase.
• Semantic emphasis — LLM-pass per phrase: “which 1–2 words carry the meaning?” Highlight those. Don’t default to stressed syllables or loudest words.
• Silence honor — if speaker pauses 1.5s+ for rhetorical effect, don’t back-fill with lingering prior caption. Let silence breathe.
• Quote-sensing — “he said, quote…” or air quotes → italic (one of the allowed italic exceptions).
• No [laughs] / [sighs] — those are accessibility captions. For aesthetic captions, they pollute the frame.

Self-critique checklist (run before rendering)

The agent’s own pre-render pass. Flag violations:

• Not done: Does the caption or subject read first? (Rule #1)
• Not done: Any hard white pill boxes? (Rule #3 — banned on cinematic)
• Not done: Still lower-third for a close-up? (Rule #4)
• Not done: More than 2 weights or more than 1 font family? (Rule #5)
• Not done: Italic used for emphasis? (Rule #8 — banned)
• Not done: More than 1 saturated accent color? (Rule #9)
• Not done: Letter-spacing or filter:blur animating? (Rule #10 — banned)
• Not done: Stagger same for documentary and vlog? (Rule #12)
• Not done: Every group emphasized? (Rule #13)
• Not done: 30s+ with no rhythm-break? (Rule #14)
• Not done: Displaying every filler word? (Rule #15)
• Not done: Caption crossing breath-break? (Rule #16)
• Not done: Caption in platform UI zone? (Rule #17)

Any “yes” to a violation question → regenerate the affected segment.