Grand Diomande Research · Full HTML Reader

Stillness Catalog — Content System Derived from @nouses_kou

Where @nouses_kou inscribes the void in Noh-Japan-AR, you inscribe N'Ko-Mande-Sensing. Same architecture, different ancestry.

Embodied Trajectory Systems research note experiment writeup candidate score 32 .md

Full Public Reader

Stillness Catalog — Content System Derived from @nouses_kou

Not a copy. The grammar borrowed, the substance native.

Where @nouses_kou inscribes the void in Noh-Japan-AR, you inscribe
N'Ko-Mande-Sensing. Same architecture, different ancestry.

---

I. THE DNA — what makes the grammar work

Eight invariants. Every post in this system obeys all eight. The
discipline IS the brand.

#InvariantWhy it worksYour translation
1Fixed-frame, single-take, 17-30sNo cuts means the viewer cannot escape. Attention IS the medium.The locked frame mirrors the act of sensing. LUME is a stillness product.
2No face, no voiceThe author disappears. Subject becomes object.You don't have to be the protagonist. The hand, the leaf, the room, the N'Ko glyph is.
3One sustained drone, never trending audioSound is felt, not selected for the algorithm.You can generate drones FROM the brain. The body's 128D vector synthesizes its own score.
4Overlay markers — tracker IDs, point clouds, pseudo-codeThe implementation IS the visual.Your 33 BlazePose joints, your 128D slots, your phase counter ARE the aesthetic. Show them.
5Caption: glyph string → classical quote → English gloss → sign-offText is a poem with ritual closing.N'Ko glyphs → Maninka/Bambara proverb → English gloss → `ߥߟߊ` (Walaha)
6Three hashtags, locked, never rotatedHashtags are a flag, not bait.Pick three. Never deviate.
7The catalog is the brandOne viral post does nothing. 20 posts in lockstep grammar build a gravity well.Ship the series; don't chase the single hit.
8No CTAAsking for the click cheapens the artifact.The work stands. People come because the room is built.

---

II. THE THREE SERIES — full shoot plans

Each series obeys all 8 invariants. Each post in a series obeys the
same sub-grammar. Variations come from substrate, not structure.

---

SERIES A — Hands That Know (5 posts)

Concept 1 expanded. The hand as instrument of inscription.

Frame
- 22 s, 4K vertical (1080×1920), 30 fps, locked
- Plain wall background. Texture matters: limewash, raw concrete, brown paper, weathered wood, linen
- One light source, one shadow, no fill
- Hand enters from frame-right, opens and closes three times, exits frame-right
- Distance: arm fills lower-third only. Wall is the protagonist

Overlay
- Single neon-orange tracker line, 1px wide, smoothed centroid of wrist+middle-knuckle (MediaPipe Pose landmarks 15+19 averaged)
- The line accumulates the trail of the hand's path with a 1.5s exponential decay
- One N'Ko character pulses in upper-third at 0.8 Hz (one breath per 1.25 s)
- Character changes per post in the series

Audio
- 200 Hz sine wave drone, -18 dBFS
- Layered with faint room hum recorded on iPhone in actual venue
- Optional: one glitch tick at the moment the hand closes fully (sample: pencil tap on paper)

Caption template

⟶ ߛ ⟵

「手は知っている」  the hand knows

ߥߟߊ

(Replace `ߛ` with that post's character. Replace Japanese with a
Maninka or Bambara proverb. The gloss is one English line.)

Hashtag triad (lock)

#nko #computationalchoreography #lume

The 5 posts in the series

#N'Ko characterSubstrateProverb (Maninka)Gloss
A1ߛ (sa)Limewashed wall"Bolo dɔ kelen tɛ se ka kungo bila"one hand cannot tie a bundle
A2ߡ (ma)Raw concrete"Bolo gbɛ̌ ka di ni bolo kojugu ye"a clean hand pleases more than a strong one
A3ߞ (ka)Brown craft paper"Bolokɔnin tɛ se ka ji to"the small finger cannot hold the water
A4ߓ (ba)Linen sheet"Bolo fila bɛ ɲɔgɔn ko"two hands wash each other
A5ߟ (la)Weathered cedar"Bolokurun tɛ kuma"the closed hand does not speak

---

SERIES B — Inscriptions (7 posts)

Concept 2 expanded. N'Ko writing itself onto the natural world.

Frame
- 28 s, 4K vertical, 30 fps, locked
- Macro shot, 1:1 or 4:5 cropped to vertical
- Subject fills 70
- Natural backlight (golden hour OR diffused indoor through paper)
- No human in shot. Ever.

Overlay
- N'Ko proverb writes itself character-by-character along a natural curve in the substrate (leaf vein, water surface, sand groove, wood grain)
- Per-character delay: 1.2 s — so a 6-character proverb writes over 7.2 s, leaving ~20 s of stillness for the finished inscription to breathe
- Character color: bone-white (#F2EEE5) with 0.5px black drop shadow for legibility on any substrate
- After the last character lands, a faint 1px line connects all characters along the chosen curve

Audio
- Field recording, no music
- Recorded in the same location as the shoot (or a complementary one): wind, distant water, distant birds, room tone
- Mixed flat at -22 dBFS, no compression, one micro-fade at the very end

Caption template

ߞߌ ߟߏ ߞߊ ߓߐ

"kɛlɛ tɛ kɛ ka bɔ"  before the war, peace must depart

ߥߟߊ

(One proverb per post. The N'Ko script in the caption mirrors what
writes itself in-video.)

Hashtag triad (same lock)

The 7 posts in the series — one per N'Ko vowel + organic substrate

#N'Ko vowelSubstrateLightField recording
B1ߊ (a)Monstera leaf, backlitGolden hour, west windowWind through screen door
B2ߋ (e)River stone, wetOvercast noonStream, mid-distance
B3ߌ (i)Pool of still waterDawn, low angleSingle bird, far
B4ߍ (è)Dune sandHigh sun, rakingWind across sand
B5ߎ (u)Reclaimed oakWindow light, indirectEmpty room tone
B6ߏ (o)Fresh snowDiffused overcastSnow muffling distant traffic
B7ߐ (ɔ)Old tree barkDappled forest lightForest mid-distance

The seven posts together form a complete vowel pass — the alphabet
inscribed onto the world.

---

SERIES C — Rooms Remember (4 posts)

Concept 3 expanded. The architecture of presence after absence.

Frame
- 30 s, 4K vertical OR cinematic 16:9 (your call, but pick one and stay)
- Locked wide of an empty interior space
- Spaces: a bar without patrons, a hallway, a kitchen pre-service, a stairwell, a chapel
- One light source, late afternoon or evening, warm
- No human visible in the live frame

Overlay — the technical heart
This is the post that justifies the whole pipeline.

Pre-shoot: record yourself moving through the empty space for 30 s.
Process the recording through your Femto + MediaPipe pipeline. Extract
the 33-joint trails.

In post: render those trails as faint, slowly-blooming ghost traces on
the walls of the locked live frame. The trace fades in over the 30 s
duration. By the last frame, the ghost has fully written itself.

Color: warm white, low opacity (10-15
Spatial coherence: the trace should appear at the correct depth (use
the depth channel from your Femto recording to place trails at the
right wall, not floating in mid-air)

Audio
- 120 Hz sustained drone, generated by `echelon-bar` reading the
pre-shoot JSONL through cc-brain → dynamics_128 → simple subtractive
synth (5 voices, modulated by latent.norm)
- One micro-glitch every 4 seconds: a tiny burst of high-frequency
noise gated to when periodicity > 0.5 in the recorded vector
- Total length matches the 30 s clip exactly

Caption template

⌛ ◯ ⌛

the room remembers

ߥߟߊ

The 4 posts in the series

#RoomTime of dayWhose traceDrone fundamental
C1Empty bar (Plaisir)Late afternoonYours120 Hz
C2Apartment hallwayDuskYours110 Hz
C3StairwellNight, single bulbA friend's130 Hz
C4Kitchen, pre-serviceMid-morningYours100 Hz

The four together: one full day of rooms remembering one full week
of bodies.

---

III. SERIES D (RESERVE) — Body As Instrument (deferred)

If A+B+C land well, the natural fourth: a series where YOUR body
moves and the visualization IS the LUME 128D vector projected onto
the wall behind you in real time. But that's only post-D3 NSSM and
post-D4 reboot survival. Don't ship D until C is published.

---

IV. PRODUCTION PIPELINE

What you build into the repo to make this trivially shootable.

Three small tools to ship in `core/audio-media/cc-echelon/tools/content/`

1. `render-overlay.py` — takes a `.jsonl` pose recording + a `.mov`
live shot, emits a `.mov` with the neon-tracker overlay composited.
Single command:

   python render-overlay.py --jsonl hand.jsonl --video hand.mov \
     --color "#FF6A1A" --width 1 --decay 1.5 --nko "ߛ" \
     --out hand-final.mov

Internally: ffmpeg + opencv. Reads BlazePose 33-joint world coords
from JSONL, projects to image space using known camera intrinsics
(we have them — Femto returns them), draws polyline with exponential
decay, overlays N'Ko at fixed position with pulsing alpha.

2. `render-typewriter-nko.py` — takes a `.mov` (Series B substrate)
+ a string of N'Ko + a curve descriptor (image-space path), emits
the same `.mov` with characters appearing one-by-one along the curve.

   python render-typewriter-nko.py --video leaf.mov \
     --nko "ߊߡߊߞߊ" --curve leaf-vein.json --delay 1.2 \
     --out leaf-final.mov

The `--curve` is a JSON list of (x, y) waypoints traced manually
in any image editor over the first frame. 5 minutes to author.

3. `synth-drone.py` — takes a `.jsonl` pose recording, emits a
`.wav` drone whose fundamental and amplitude track the body's
energy and periodicity.

   python synth-drone.py --jsonl room.jsonl --fundamental 120 \
     --voices 5 --duration 30 --glitch-on-periodicity 0.5 \
     --out room-drone.wav

Internally: reads JSONL → feeds each frame through cc-brain's
`update_femto_skeleton` + `dynamics_128` → 5 sine voices whose
detuning is modulated by latent.norm → glitch burst when
periodicity threshold crosses. Lives next to `audio-engine` crate.

These three tools turn each shoot from "needs a video editor" to "one
command per post." Estimated dev time: 3-5 hours total. Worth shipping
BEFORE the first Series A post so the catalog can flow.

What you already have (no new build)

  • iPhone (4K vertical, 30 fps)
  • Tripod or stable surface
  • Mac1 with full LUME stack + MediaPipe + Echelon + cc-brain (all green)
  • Femto Bolt OR Femto Mega (for Series C pre-shoot pose capture)
  • N'Ko expertise (you ARE the source)
  • A wall, a leaf, a room (the world supplies)

What you do NOT need

  • After Effects (the three tools above replace it)
  • A DSLR (iPhone is sufficient at this aesthetic scale)
  • Studio lighting (natural light is the look)
  • A voice-over rig (no voice)
  • Trending audio research (no trending audio)

---

V. CADENCE — how the catalog assembles

16 posts across 3 series, rotating ABCABCABC. One post every 2-3 days.
Six weeks to a complete catalog.

Week 1: A1, B1, C1
Week 2: A2, B2, C2
Week 3: A3, B3, C3
Week 4: A4, B4, C4
Week 5: A5, B5     (Series A complete, Series B at 5/7)
Week 6: B6, B7      (Series B complete, full catalog of 16)

Post times: don't optimize. Pick 11:11 AM ET local and never deviate.
The reliability is a signal.

Account: @diomandee on Instagram (your existing handle). Don't
cross-post to TikTok yet — let IG build its own gravity well first.
Cross-post starts at Week 5 when the catalog has critical mass.

---

VI. WHAT YOU DO NOT BORROW

Three signatures that are HIS, that would feel imitative if you copied:

1. The "Thank you very much, sincerely." sign-off. That's deeply
Japanese and personal to him. Your sign-off is `ߥߟߊ` (Walaha) —
Maninka/N'Ko for "peace upon it" or "thanks." Use yours.
2. Noh quotes. Those are his literary universe. Yours is N'Ko
proverbs, Maninka oral tradition, Souleymane Kanté. Borrow nothing
from his quote register.
3. The exact neon-green tracker palette. His green is a
TouchDesigner default. Your palette is warm — orange for fire/voice
(ߛ), bone-white for inscription, never the cool green. Different
ancestry, different temperature.

---

VII. THE OPENING POST — sequence to ship A1 this week

Concrete steps. Do these in order. Ship by Sunday.

1. Today, 15 min: walk your apartment, find the limewashed-or-
plain wall. Mark the tripod position with tape. Note the time of
best light (when the shadow is hardest).

2. Today, 30 min: ship `render-overlay.py` v0 to the repo
(`core/audio-media/cc-echelon/tools/content/`). Don't make it
perfect, make it work for one hand video.

3. Tomorrow, light hour: shoot 5 takes of the hand opening +
closing three times. Best take wins. 22 s exact.

4. Tomorrow, 1 hour: run the take through `render-overlay.py`
with `--color "#FF6A1A" --nko "ߛ" --width 1 --decay 1.5`.
Audio: record the 200 Hz drone in any DAW, layer your bedroom
room-tone underneath.

5. Tomorrow, 10 min: caption + post. Don't second-guess. The
catalog needs its first stone.

6. +2 days: B1 (monstera leaf, golden hour) following the
exact same pipeline with `render-typewriter-nko.py`.

7. +4 days: C1 (empty bar at Plaisir, late afternoon) following
`synth-drone.py` for the audio.

After A1+B1+C1 you have proof-of-system. The remaining 13 posts are
execution, not invention.

---

VIII. ONE-LINE COMPRESSION

> Locked frame. No voice. One drone. Glyph overlay. Caption is a
> proverb in N'Ko with an English gloss and `ߥߟߊ` to close.
> Three hashtags. No CTA. Sixteen posts. Six weeks. Walaha.

That sentence is the entire system. Print it. Put it above the
tripod.

Promotion Decision

Attach run IDs, datasets, metrics, and reproduction commands.

Source Anchor

lume-content/nouses-kou-derived-content-plan.md

Detected Structure

Method · Evaluation · Code Anchors · Architecture