The Cloud
🔭

AARON Totem — the 360° capture pillar

The CloudThe Cloud

AARON Totem — the 360° capture pillar

🔭

A concept page to flush out the capture identity of the Totem — a mini standing pillar (DJI-Mic-3 scale) that is a 360° volumetric camera system. It documents your lived experience in real time, lets you (or an agent) edit that footage directly, and worn-or-placed it travels with you. This is the device-as-witness companion to the Totem's other role as a thin Cloud terminal.

This sits under AARON — Tablet & Totem. That page owns the Totem-as-terminal (standing/podium/civic surface). This page owns the Totem-as-capture-device. Same object family, two halves of its purpose.


The one-line pitch

A pocket-pillar that captures everything around you in 360°, holds nothing on-device, and turns lived experience into editable, AI-shapeable footage that lives in your Cloud account.

Form factor

  • Scale: mini pillar / standing totem, roughly DJI Mic 3 size — small enough to wear, hold, or set down. (Exact dimensions referenced from the video — to be filled in.)

  • Stands on a table on its own — a self-supporting 3D capture station.

  • Wearable as a necklace — pendant orientation puts the camera array at chest height for hands-free first-person documentation.

  • Held like a GoPro for active/action capture.

  • Drops into 3D composite cases / mounts — bike, helmet, tripod, dash, etc. (ties directly to AARON Skins & Kit — open-IP 3D-printed cases, accessories & the print marketplace).

Capture system

  • Four outward-facing cameras → stitched into a true 360° / volumetric field. "Meta footage" — capture the whole scene, choose the frame later.

  • Real-time documentation of everything you're experiencing — set-and-forget, ambient witness.

  • Voice-forward. Heavy voice interaction; talk to it, narrate, command, query.

  • Tiny Tamagotchi-scale screen. A small expressive display on the body — status, framing peek, a bit of personality — sized to the object, not a viewfinder.

Edit & intelligence

  • Footage streams to your Cloud account (nothing personal lives on the device — same thin-terminal principle as the rest of AARON).

  • Direct editing of the 360° capture, or hand it to an AI/agent to reframe, cut, restyle, and produce in various ways.

  • Because the raw is volumetric, the "shot" is a decision made after the fact — by you or by an agent.

Where it fits in the AARON family

  • The Tablet — large, iPad-scale personal-but-shareable surface.

  • The Totem (this) — the standing pillar; terminal and 360° capture device.

  • The large public screen — a home/public big-screen surface anyone can authenticate into.

The connecting principle — identity, not hardware

🔑

Authentication is what gets you, not hardware storage. Every AARON device holds nothing personal. You authenticate by your method of choice and your OS, apps, data, and captured footage stream in; sign out and it returns to ambient. That makes the whole fleet interoperable and shareable — pick up any device anywhere, it becomes yours, then forgets you. The Totem's camera captures to your account, not to a card in this object, which is why a borrowed or shared Totem is still safe and still yours.

Teka alignment

One shared, long-lived capture fleet instead of a camera hoarded per person — less manufactured hardware, longer lifecycles, less e-waste. Capture and computing as a utility. Help the human, help the earth.


Open questions to flush out

Exact dimensions / proportions (pull from the reference video).
Camera count & lens spec — is four final, or does volumetric stitching want more?
On-device buffer vs. pure stream — does it need any local cache for connectivity gaps, and how does that square with "holds nothing"?
Power & runtime as a wearable vs. a table station.
The Tamagotchi screen: pure status/personality, or minimal framing/playback too?
Privacy model for capturing others in shared/public space — consent signaling, recording indicator.
Relationship to the Skins & Kit mount system — which mounts ship first.



Prototype BOM — parts to build the first Totem

🛠️

Goal of this list: a maker-grade, off-the-shelf bill of materials to stand up a working DJI-Mic-3-scale capture pillar — 360°(ish) camera array + voice-forward mics + tiny status screen + cloud streaming + battery. This is the "prove the concept" build, not the final industrial design. Prices fluctuate and are omitted on purpose; treat quantities and example parts as a starting kit.

The one architecture decision that drives everything

⚠️

Four cameras can't share one camera port. A single CSI port + multiplexer board only makes one camera live at a time — fine for stills/surveillance, not for true simultaneous 360° video. So the brain choice is the whole ballgame. Three viable paths, cheapest/fastest first:

Path

Brain

Cameras live at once

Good for

Trade-off

A — MVP (recommended first build)

Raspberry Pi 5 (native CAM0 + CAM1)

2 (front/back)

Proving stream-to-Cloud, edit pipeline, mics, screen, power, enclosure

"180°×2" not a seamless sphere yet

B — Full 360°

Compute Module 4/5 on a quad-CSI carrier (e.g. Arducam / dual-IO carrier)

4

True 4-lens volumetric stitch

More integration work, bigger board, more power/heat

C — Distributed

2–4× Pi Zero 2 W, one per camera, streaming over the network

4 (1 each)

Matches the page's "feed compressed streams over a network" idea; modular

Sync + clock alignment across nodes is real engineering

Build A first. It exercises every other subsystem below; swapping in B or C later only changes the brain + camera count.

Bill of materials

#

Subsystem

Example part(s)

Qty (Path A)

Est. price range (USD)

Notes

1

Compute / brain

Raspberry Pi 5 (4–8 GB)

1

$60–80 (16 GB ~$120)

Native dual CSI = simplest 2-cam capture. Swap to CM4/CM5 + carrier for Path B (carrier adds ~$40–100).

2

Cameras

Camera Module 3 (IMX708) std + wide; or Arducam IMX519 AF / IMX477 HQ

2 (A) / 4 (B,C)

$25–40 each → $50–160

Wide-angle reduces lens count. HQ/IMX477 is the pricey end. Don't mix sensors on one multiplexer.

3

Camera cabling

15-pin → 22-pin CSI adapter FFC cables (Pi 5 small connector)

2–4

$10–30

Buy a few lengths; Pi 5 / Zero connector differs from Pi 4.

4

Microphones (the "DJI Mic" half)

I2S MEMS — ICS-43434 (high SNR) or SPH0645; INMP441 budget alt

2–3

$3–13 each → $10–40

Array of 2–3 enables direction/beamforming. Digital I2S avoids analog noise; keep bottom port facing out.

5

Status screen ("Tamagotchi" display)

1.28" round IPS LCD, GC9A01, 240×240, SPI

1

$10–20

Driven over SPI; shares GPIO with mics — plan the pin budget.

6

Orientation / stabilization

IMU — MPU-6050 (basic) or BNO085 (fused)

1

$5–30

"Which way is up" for stitching + horizon-lock; needed for pendant wear. BNO085 is the high end.

7

Power

LiPo (3.7 V 3000–5000 mAh) + UPS/boost HAT (PiSugar, Pi UPS HAT, PowerBoost)

1 set

$40–90

Size the cell to whichever runtime you test first (wearable vs. table station).

8

Connectivity (beyond Wi-Fi)

4G/5G USB modem or cellular HAT (+ SIM)

1 (optional)

$40–300

Wi-Fi is built in; add cellular only for untethered field streaming. 5G is the top of the range.

9

Storage / boot

microSD (A2, 64 GB+) — or NVMe via Pi 5 M.2 HAT

1

$10–70

Boot + connectivity-gap buffer only. Footage streams to the Cloud account, not a card.

10

Thermals

Active cooler / heatsink + fan for Pi 5 (or CM heatsink)

1

$5–10

Continuous multi-cam encode runs hot — mandatory for sustained capture.

11

Controls & indicators

Tactile button(s) + recording-indicator LED

2–3

$5–10

The LED doubles as the privacy/consent "I am recording" tell.

12

Enclosure

3D-printed pillar body + apertures + necklace / GoPro mount

1

$3–15 in filament

Printer itself is one-time tooling — see below. Ties to AARON Skins & Kit.

13

Misc / glue

USB-C cabling, JST connectors, jumper wires, standoffs, small carrier PCB

1 set

$15–40

Mic array + LED + button + screen are easier on one small carrier than flying leads.

💵

Rough per-unit electronics total (Path A, excl. optional cellular + printer): ~$250–600. Add cellular ($40–300) only if you go untethered. Path B/C add cameras + a carrier or extra Pi Zero nodes (~$15–25 each).

#### Tooling — the 3D printer (one-time)

The enclosure is tiny (DJI-Mic-3 scale), so build volume is irrelevant — even the smallest machine fits it many times over. What matters: it should be enclosed so you can print PETG / ABS / ASA for a durable worn body that tolerates heat from the electronics, and dimensionally accurate for the press-fit camera/mic apertures. FDM is the right tool; resin is the wrong one here (brittle, messy, small bed — only worth it for a non-functional show model).

Tier

Printer

Est. price (USD)

Why for the Totem

Best value (recommended)

Elegoo Centauri Carbon

~$300

Enclosed, fast, prints PETG/ABS/ASA out of the box, accurate enough for the apertures. Most capability per dollar.

Reliable workhorse / ecosystem

Bambu Lab P1S (discounted since P2S launch)

~$500–700

Enclosed CoreXY, very low failure rate, huge community, ties cleanly to a print-marketplace workflow.

Precision / premium

Prusa Original CORE One

~$950+

Best FDM dimensional accuracy + open-source; ideal once you're iterating tight tolerances for the production body.

Fit-test only (skip for final)

Bambu Lab A1 mini

~$200

Cheap and easy, but open-frame and PLA/PETG only — fine for shape mockups, not a durable worn enclosure.

Recommendation: start with the Elegoo Centauri Carbon (~$300) — it's enclosed and prints engineering materials, so the same machine carries you from first fit-test to a wearable-grade body. Step up to the Prusa CORE One only when you're refining final tolerances.

Suggested first-build sequence

  1. Bench the brain + 2 cameras on a Pi 5, get both CAM0/CAM1 streams running at once.

  2. Add the I2S mic array and confirm synced audio+video capture.

  3. Stand up the Cloud stream + buffer — footage off-device, microSD only as a connectivity-gap cache.

  4. Wire the round screen, LED, button, IMU onto one small carrier.

  5. Power it off the LiPo/UPS and measure real runtime in pendant vs. table mode.

  6. Print the pillar enclosure and integrate. Only then decide whether to jump to the 4-camera CM-based brain (Path B) for true 360°.

Open hardware questions this BOM surfaces (feed back into the list above)

Camera count for true 360°: is 4 wide lenses enough, or does seamless volumetric stitching want more / a different FOV mix?
Brain path: commit to Pi 5 dual-cam MVP now, but decide CM4/CM5-quad vs. distributed Zero-node for the 360° build.
Buffer policy: how big a local cache for connectivity gaps without violating "holds nothing personal"?
Mic array geometry: 2 vs. 3 mics, and spacing, for usable direction/beamforming.
Power budget: continuous multi-cam encode + cellular is hungry — match cell size + cooling to target runtime.
Recording indicator: spec the LED/screen behavior as the consent/privacy tell for capturing others.



Shopping checklist — Path A starter kit (Amazon)

🛒

Tagged Amazon links for the recommended Pi 5 dual-camera build + printer. As an Amazon Associate, Ora earns from qualifying purchases. Links are tagged search links — pick the best-reviewed seller on each. Prices on the Pi + cameras move around; confirm at checkout. Rough total: ~$300 printer + ~$300–400 parts for one unit.

Core kit (from the cards)

Raspberry Pi 5 (8 GB) — the brain; native dual CSI for front/back capture → search
Raspberry Pi Camera Module 3 — buy 2 (consider Wide / 120°)search
CSI camera cable 15-pin → 22-pin (Pi 5 adapter) — get a 2-packsearch
INMP441 I2S MEMS microphone — buy 2–3 for an array (ICS-43434 = higher-SNR upgrade) → search
GC9A01 1.28" round IPS display (240×240, SPI) — the Tamagotchi status screen → search
Raspberry Pi 5 Active Cooler — mandatory for sustained multi-cam encode → search
Raspberry Pi 5 UPS HAT + lithium battery — untethers for wearable use → search
Elegoo Centauri Carbon 3D printer — enclosed, prints PETG/ABS/ASA → search

Commodity extras (any decent listing, ~$25–45 total)

microSD card, 64 GB+ (A2) — boot + connectivity-gap buffer only → search
IMU — MPU-6050 (basic) or BNO085 (fused) — orientation / horizon-lock → search
Tactile buttons + LEDs — record/wake + the recording-indicator tell → search
Dupont jumper wire kit — wiring the carrier → search
PETG filament, 1 kg (~$20) — durable enclosure material → search



Form factor reality check — how big the Path A build actually is

📏

The off-the-shelf Pi 5 stack does NOT hit DJI-Mic-3 scale. It's the bench prototype that proves the pipeline; pendant-scale is a later generation that needs custom integration, not stacked HATs.

Estimated enclosure: ~90–100 × 65–75 × 60–80 mm, ~300–450 g (with battery + printed PETG body).

The footprint is locked by the Raspberry Pi 5 board (85 × 56 mm — standard Pi size). Height piles up because three things compete for the vertical axis:

Layer

Adds (height)

UPS HAT + battery

~15 mm (pouch LiPo) / ~22–25 mm (18650 cells)

Pi 5 board + connectors

~16–20 mm

Active Cooler (clips on top)

~10 mm

Cameras (×2), round display, mics, IMU, and buttons mount on the faces and mostly recess in — dimensionally minor next to the core stack.

vs. the anchor: DJI Mic 3 transmitter = 28.7 × 28.3 × 16.3 mm, 16 g. This build is ~3× longer, ~2.5× wider, ~4× taller, ~30× the volume and ~20–25× the weight. Tangibly: about three decks of playing cards stacked, or a small coffee mug — a "hold it / set on a table / clip to a bag" object, not a collar pendant.

Shrink path (when size becomes the priority)

  • Smaller brain — Pi Zero 2 W (65 × 30 mm, ~half the footprint) or an ESP32-class MCU. Cost: loses dual native CSI + real-time stitching horsepower → pushes you to Path C (networked Zero nodes per camera) or a custom board.

  • Pouch LiPo instead of 18650s + a custom carrier PCB instead of stacked HATs — can roughly halve the height on its own.

🎯

Takeaway: prove the concept at coffee-mug scale on Path A, then treat "DJI-Mic-3 scale" as a dedicated industrial-design / custom-PCB milestone — it won't fall out of the off-the-shelf stack.



Housing CAD & verified blueprints

🔭

Live tool: 3D — Totem Housing CAD — the new 3D lens in Studio (alongside App / Music / Video). Parametric enclosure model with orbit + exploded view, dimensioned blueprint views (front / side / top), live fit-check, and STL + SVG export.

Verified external envelope (at default tolerances): 96.0 × 67.0 × 62.0 mm. Internal cavity 91 × 62 × 57 mm. Defaults: wall 2.5 mm, clearance 3 mm, layer gap 3 mm, lid 2.5 mm.

What the double-/triple-check caught:

90 mm external was too small. 2.5 mm walls leave 85 mm internal = exactly the Pi 5 width, zero clearance — unbuildable. Verified minimum is ~96 mm to keep a real 3 mm clearance. (Earlier form-factor estimate corrected.)
It's a brick, not a pillar (~1.5 : 1). Because HATs stack vertically, the footprint is locked at the Pi 5's 85 × 56 mm and the stack is short. The "360° capture pillar" silhouette is not achievable on the Pi 5 path — it needs the smaller-brain path (Pi Zero 2 W / MCU).
~30× the volume of a DJI Mic 3 — the tool shows this live, consistent with the form-factor reality check above.
Shell exports as a printable tray + lid (slab construction). Good as a first STL to print; refine camera/mic ports, screen bezel, and standoff bosses in full CAD before a final body.

Open CAD tasks (next pass)

Add real port cutouts (2× camera, mic grille, USB-C, round-display bezel) — currently shown as placements, not subtracted from the shell.
Add mounting bosses / standoffs for the Pi 5 (M2.5) and the UPS HAT.
Decide lid fixing (snap-fit vs. M2.5 screws + heat-set inserts).
Model the necklace loop / GoPro mount from the Skins & Kit page.
Re-run the fit-check against the pillar (Zero-node) build once that BOM exists.
The Cloud