reel · 12-15s · 2026-05-25approvedfact-checkoffer-tied

Ad-hoc Brief, Bipartite research-authority reel on the relationship a woman has with her own ADHD

blockers: fact-check has unsupported claim

No images yet. Visual director writes the brief, then image-generation-protocol renders the PNGs via Higgsfield.

No `## Text on video` section in the draft — only reels need this.

Caption25 sentences · 264w

Researchers spent nearly three decades studying what actually changes when a late-diagnosed woman finally understands her ADHD, and the finding was not the one anyone expected.

It was not a new strategy. It was not a better planner. It was not medication, though that helps too.

The shift was internal.

A 2014 review on missed and late ADHD diagnosis in women named what the literature had been circling for years. Women receive the ADHD label, on average, years later than their male peers. By the time the diagnosis arrives, the brain has spent years collecting evidence against itself.

The forgetfulness gets filed as carelessness. The lateness gets filed as disrespect. The half-finished projects get filed as proof.

Each failure feeds the self-concept. The self-concept makes the next failure harder to recover from. The loop tightens.

For many late-diagnosed adults the loop becomes the harder problem. The ADHD is the wiring. The shame is what makes the wiring unworkable.

The pivot is quiet when it happens. The brain stops being the villain in the story and becomes the subject of study. The wiring has a name. The name has a manual. And the interventions that slid off for twenty years suddenly hold, because curiosity is a propellant that shame was never going to be.

If the shame loop is the part that has never responded to a single planner or productivity book, Phase 1: Understand of The ADHD Playbook, inside the Complete System That Works With Your ADHD Brain bundle, was built for this exact internal pivot. Link in bio.

adhdwomen, latediagnosedadhd, adhdshame, adhdrelationship, adhddiagnosis

Rhythm legendwhy some words are highlighted

Sentences over 18 words are tagged via tooltip. Three in a row gets a soft amber underline — the brand voice prefers rhythm variation, not monotone walls of long lines.

Visual brief

Visual brief: Bipartite research-authority reel on the relationship a woman has with her own ADHD

Format: reel Generated: 2026-05-25T00:00:00Z Source draft: drafts/2026-05-25/relationship-with-adhd.md Source brief: briefs/ad-hoc/2026-05-25-relationship-with-adhd.md Length: 14 seconds (within the brief's 12-15s window; aim 14s, hard cap 15s) Shot count: 1 (single continuous take, no cuts)

Overall visual concept

One continuous macro close-up of a single human eye filling 60-70% of a 9:16 vertical frame, dark moody key with a single warm light from camera-left. The pupil dilates very slowly across the run, the eye blinks once near the midpoint, and the reflection inside the iris sharpens from an indistinct gray smear into a small visible square of warm light by the closing frames. A bipartite static text overlay sits over the frame from frame 1 to the last frame, never animates, never changes. The whole reel reads as a quiet interior shift made visible.

Brand palette (hex codes in use)

The reel sits inside the brand's warm-and-grounded register, but pushed toward the moody end of it (this is a dark-key scene, not a cream-paper scene). Active hexes:

Skin / shadow base: warm umber drift toward brand deep brown #3D2818 in shadow, lifting to charcoal warmth #2A1F18 in the deepest unlit areas. No true black anywhere.
Warm key light on skin: sits between rust #B86C50 and cream #F4ECDC in the highlight roll-off. Treat as practical-lamp warm, around 2700-3000K in feel.
Eye reflection (final state): a small square of warm light that reads in the #F4ECDC / #FAF3E2 band. Cool reflection highlights on the wet of the eye drift toward dusty blue #7C95A6 to keep the eye legible against the warm skin.
Iris: writer/director discretion on iris color (brown, hazel, green, blue all acceptable). Whatever the iris is, grade it so it sits inside the brand palette, not outside it. Avoid saturated jewel tones.
Overlay text: cream #F4ECDC over the dark frame, NOT pure white. See typography.
Avoid (per protocol section 0): pure white, pure black, neon anything, gradients, hard drop shadows.

Typography (overlay only)

The bipartite overlay is the entire on-screen text for the reel. No lower thirds, no captions appearing mid-reel, no end card.

Typeface: Recoleta (primary) or Lora bold (fallback) per the brand display stack. Both lines use the SAME typeface and weight. No mixing.
Weight: semi-bold (Recoleta SemiBold) or bold (Lora Bold). Lean semi-bold if Recoleta is available; the moody contrast already does the heavy lifting.
Case: sentence case on both lines. This is emotional-anchoring content, not educational headers, so sentence case per protocol section 0.
Color: cream #F4ECDC for both lines. Not pure white. The cream warms into the scene; white would punch out of brand.
Alignment: center-aligned horizontally on the 9:16 frame.
Line wrap: see the line-break section below. Both lines wrap to multiple visual rows; do not force a single row.
Letter-spacing (kerning): default Recoleta tracking. Do NOT widen-track or condense. If using Lora, the same: default tracking only.
Line-height: 1.15 to 1.20 within each text block (tight enough to read as one paragraph, loose enough to breathe).
Gap between top block and bottom block: roughly 18-22% of the frame height. This is the visual breath that separates the authority-timeframe from the reframe-twist. The eye sits in the gap.
Size (relative to a 1080x1920 9:16 canvas):
- Top line: ~46-50pt. It is 15 words and wraps to four visual rows. The total top block is taller than the bottom block by design.
- Bottom line: ~52-58pt. Slightly larger than the top line to give the reframe-twist visual primacy. It is 10 words and wraps to two or three rows.
NO text effects: no drop shadow, no outline, no glow, no gradient fill. If legibility against the eye is borderline, lift the underlying frame's exposure by 1-2% in those zones rather than adding a text effect.

Safe-zone placement (9:16, 1080x1920 canvas)

Instagram crops roughly the top ~14% (top bar, profile chrome on the chase frame) and the bottom ~24% (caption, like/comment/share/save column overlap, audio attribution). Safe zone for hook text is the vertical band from ~14% to ~76% of frame height.

Top block (top line, 15 words, 4 visual rows): anchored from ~16% to ~36% of frame height. Centered horizontally, with ~6% horizontal margin on either side (so text body is centered in ~88% of the frame width).
Eye / focal point: the iris and pupil sit centered horizontally and roughly at ~50% of frame height (slightly above true center to compensate for the bottom block sitting heavier). The eye is the optical anchor in the gap between the two text blocks.
Bottom block (bottom line, 10 words, 2-3 visual rows): anchored from ~62% to ~74% of frame height. Centered horizontally with the same ~6% margin.
Below 76%: intentionally clear. This is where the Instagram caption preview and the engagement column overlap on most devices. Nothing critical lives there.

Cinematography

Shot type: extreme close-up (macro) of one human eye. The eye fills 60-70% of the frame height. Eyebrow optionally visible at the top edge of frame, upper cheek visible at the bottom edge. Hair, ear, full face are NOT in frame.
Lens / look: macro lens equivalent, 90mm to 105mm full-frame feel. Shallow depth of field, focus plane locked on the iris. Lashes pick up the key light and are in acceptably-soft focus, not razor sharp.
Angle: straight-on, eye-level, parallel to the subject's gaze. The subject looks just past the lens, not directly into it, so the reflection geometry lands clean and the gaze does not feel confrontational.
Framing for 9:16 (vertical): the iris sits center-horizontal, slightly above true vertical center (around 48-50% of frame height for the pupil center). Headroom above the eyebrow is minimal; the upper cheek roll fills the lower third. This leaves clean negative space for the bipartite overlay above and below.
Lighting: single warm key light from camera-left, roughly 30-45 degrees off-axis from the subject's nose, slightly above eye level (catchlight sits in the upper portion of the iris). Practical-lamp warmth, around 2700-3000K. No fill, no rim, no background light. The unlit side of the face falls off into warm shadow, never into pure black. A subtle bounce (white card or warm reflector) on camera-right is acceptable only if the unlit shadow is reading muddy.
Color grade: warm shadows (amber drift in the umber #3D2818 to #2A1F18 range on the skin), cool highlights in the wet of the eye (dusty-blue lift in the catchlight and the reflection's specular). Mid-contrast curve, not crushed. Faint film grain (5-8% intensity), no digital sharpening artifacts. Slightly desaturated overall, in line with the brand photography rule.
Background: there is no background. The frame is filled by the eye and the surrounding skin. The "background" is the unlit warm-shadow falloff on the camera-right side of the face.

Motion notes

This is the only motion in the reel. Nothing else moves. No camera movement.

t = 0.0s: opening frame. Pupil at its widest natural resting state for the ambient lighting (assume moderately dilated, ~5-6mm visible). Reflection in the iris is present but indistinct, reads as a faint gray smear with no resolvable shape. Overlay visible.
t = 0.0s to ~7.5s: pupil contracts very slowly (suggesting the brain registering more light / more clarity over time). Reflection begins to resolve, but at 7.5s it is still ambiguous, just a brighter patch with edges starting to form. Eye is steady, no micro-saccades visible (subject holds gaze).
t = ~7.5s to ~8.5s: one slow blink. Slower than a natural blink. The lid closes over ~0.35s, holds for ~0.15s, opens over ~0.50s. Total blink event ~1 second. After the blink, the eye is held steady again.
t = ~8.5s to ~14.0s: pupil continues to contract slowly, reaching its smallest state at the end (suggesting the light has come on, in the metaphorical sense the brief names). Reflection sharpens through this window, resolving by ~t=12s into a small but clearly visible square of warm light (window or lamp shape). From t=12s to t=14s the reflection holds at maximum clarity, no further change.
Reflection sharpen curve: non-linear. Most of the resolution happens in the back half (t=8s onward), so the change is barely perceptible early and unmistakable late. Use an ease-in curve, not linear. The viewer should notice the shift without being able to point to the moment it happened.
Pupil dilation curve: linear or very slight ease, slow enough that no single frame reads as movement. Net contraction across the run is roughly 2-3mm of apparent diameter.
Hard rule: no camera movement (no push-in, no rack focus, no zoom). The frame is locked. The motion is entirely inside the eye.

Audio direction

Per the writer's audio note in the draft, strictly enforced:

No voiceover. Zero spoken words.
No music. No trending audio, no licensed track, no original score.
Ambient room tone only. The faintest possible interior room tone (the kind of silence that has weight, not the kind that has hiss). Treat this as a near-silent reel where the macro shot does the work.
Optional bed: a single sub-40 dB warm drone is acceptable if total silence flags the algorithm or reads as muted-by-accident to the viewer. If used: a low warm pad, no melody, no progression, no swell, no rhythmic element. Below the perceptual threshold for "music." If unsure, leave it silent.
Captions / subtitles: not needed. The bipartite overlay IS the on-screen text. Do NOT auto-generate captions over the bottom of the frame.
Editor / post: if Andrii is on Veo 3.1 native audio generation, set the audio prompt to "near-silent ambient room tone, no music, no voice." If on a model without native audio (Runway, Kling, Sora), strip the model's default audio and lay the room tone in CapCut or Premiere in post.

Overlay rendering specs

The overlay is the single most load-bearing element after the eye itself. Hard rules:

State: static. Both lines visible from frame 1, both lines remain unchanged through the last frame. No fade-in, no fade-out, no kinetic typography, no word-by-word reveal, no highlight, no scale animation.
Composition: both blocks composited on top of the live macro plate in post (CapCut, Premiere, After Effects). Do NOT bake text into the AI video generation; render the AI clip clean, then comp the overlay on top in the edit. This guarantees the type is crisp at any platform compression.
Top line verbatim (15 words): Researchers spent 29 years studying what shifts when a late-diagnosed woman finally understands her ADHD
Bottom line verbatim (10 words): the relationship with the diagnosis matters more than the diagnosis

Line breaks

The exact line breaks are designer latitude within the constraint that the visual reads as a paragraph block, not a tight column or a ragged stack. Recommended line breaks for legibility at thumbnail size:

Top block (top line, 4 visual rows):

AI video prompt

Researchers spent 29 years
studying what shifts
when a late-diagnosed woman
finally understands her ADHD

Bottom block (bottom line, 2 visual rows):

AI video prompt

the relationship with the diagnosis
matters more than the diagnosis

Acceptable alternative for the bottom block, if 2 rows feels too dense at 52-58pt and pushes the bottom block past the safe-zone floor:

AI video prompt

the relationship with the diagnosis
matters more than
the diagnosis

Either bottom-block treatment is fine; pick the one that reads cleanest at the actual rendered size on a phone screen. Do NOT break in the middle of a noun phrase (e.g., do not break "29 / years" or "late-diagnosed / woman").

Kerning, leading, alignment

Kerning: default Recoleta (or Lora) tracking. No manual letter-spacing.
Leading (line-height inside each block): 1.15 to 1.20.
Horizontal alignment: center on the frame's vertical centerline.
Vertical alignment: top block top-anchored at ~16% frame height; bottom block top-anchored at ~62% frame height. See safe-zone section above.
No widows. If the chosen line-break treatment leaves a single word on its own row, rebreak.

Punctuation

The top line ends without a period. The bottom line ends without a period. Both are stylistic continuations, not standalone sentences. Match the writer's draft exactly. Do not add punctuation. (No em dashes anywhere, per brand banlist; the bipartite already reads as a single composed thought.)

Designer latitude vs. fixed

Fixed (do not change without writer/director sign-off):

Single continuous shot, no cuts.
Macro close-up of one eye, 60-70% of frame.
Warm key from camera-left, dark moody falloff, no fill / rim / background light.
Pupil dilation across the run.
One slow blink at ~t=8s.
Reflection sharpens from indistinct gray to a small square of warm light by the end.
Bipartite overlay, static from frame 1, both lines visible at all times.
Overlay text verbatim per the draft.
Recoleta or Lora bold display face, sentence case, cream #F4ECDC color.
9:16 vertical, 14s target (12-15s acceptable range).
No music, no voiceover, near-silent audio bed.

Designer latitude:

Iris color of the subject.
Specific subject (any human eye reading as adult feminine-presenting; brief allows but does not require this).
The shape of the final reflection (window square vs. lamp rectangle vs. soft warm rectangle, as long as it reads as "warm light source" by the end).
The choice between 2-row and 3-row bottom-block line break.
Whether a sub-40 dB warm drone sits under the room tone, or whether the bed is fully silent.
Faint film-grain intensity within the 5-8% band.
Subtle bounce on camera-right shadow side, if needed to keep shadow falloff readable.

AI video prompt (model-agnostic, single block)

This prompt is for downstream text-to-video generation (Veo 3.1, Runway Gen-4.5, Kling 2.0, Sora 2, or any future model). The bipartite overlay text is rendered in post on top of the clean clip, not generated inside the clip. Per protocol: NO on-screen text instruction, NO brand name in the prompt.

Extreme macro close-up of one human adult eye, holding steady on the iris, the entire reel an interior shift made visible through a single uninterrupted take. The camera is locked, static, no movement, no zoom, no rack focus, framing the eye at eye level, the iris centered horizontally and slightly above the vertical midpoint of a 9:16 vertical frame, the eye filling roughly 60 to 70 percent of the frame height with eyebrow at the top edge and upper cheek at the bottom edge. The subject looks just past the lens; the pupil contracts very slowly across the full duration, the eyelid completes one slow deliberate blink near the eight-second mark, and the reflection inside the iris begins as a faint indistinct gray smear and gradually resolves over the back half of the clip into a small visible square of warm light, like a window or a lamp coming into focus. The setting is a quiet dark interior, no background visible behind the face, the unlit side of the face falling off into warm shadow rather than black. Lighting is a single warm practical-lamp key from camera-left around two thousand seven hundred to three thousand kelvin, hitting the lashes and the upper iris, no fill light, no rim, the unlit side rolling into warm amber shadow; the color grade is moody with warm umber shadows on skin, cool dusty-blue specular in the wet of the eye, slightly desaturated overall with faint film grain, mid-contrast curve, no crushed blacks. Audio: near-silent ambient room tone, no music, no voiceover, no foley, optionally a single sub-audible warm low drone under the room tone. 9:16 vertical, 14 seconds.

Designer notes

No open questions, asset is fully specified above. The overlay text is composited in post over the clean AI-generated (or live-shot) macro plate; do not bake the text into the generated clip. If the chosen model returns a clip with intrusive default music or generated voice, strip the audio track in post and lay the near-silent room tone bed under the visual. Reels skip the image-generation step (no per-slide assets), so this brief terminates here.