Goal: design the look, feel and printable enclosure of the final safety alarm around our real components. The key principle — there are two separate design jobs, and most AI "3D" tools only do one. Mixing them up is the #1 mistake: the pretty AI-generated meshes will not fit the button, LED, battery or PCB.
Shape language, proportions, colour, materials — "does it look like a desirable product." This is what Aaron asked to "have a look at."
Output: renders & concept images to agree the form.
Dimensionally accurate, with button cut-out, LED lens hole, USB-C port, battery bay, PCB standoffs, snap-fits, correct wall thickness.
Output: STL/3MF you actually print and Charles can refine.
AI mesh generators (Meshy / Tripo / Rodin) produce visual models, not engineering geometry. Brilliant for Layer 1, useless for Layer 2 — they don't hold real dimensions and won't fit our components. The enclosure that gets printed must come from parametric CAD.
Layer 1 — concept / look & feel
| Tool | Best for | Price | Notes |
|---|---|---|---|
| Vizcom | Sketch → photorealistic, client-ready renders | Free + paid | Industry standard for industrial designers. Draw a rough alarm shape → polished render. Best to show Aaron. |
| Krea AI | Real-time "draw & it re-renders live" | Free + paid | Fastest for rapid colour / material / shape exploration. Great for a live play session. |
| Meshy | Text/image → spinnable 3D mockup | ~$14.50/mo | Best print-readiness of the mesh tools (97% slicer pass, Bambu/3MF), commercial rights on Pro. Treat as visual only. |
| Tripo | Cheapest / fastest image→3D | ~$12/mo | Clean topology; good budget alternative to Meshy. |
Layer 2 — the functional, printable enclosure
| Tool | Best for | Price | Notes |
|---|---|---|---|
| Fusion 360 | The actual enclosure (fit, tolerances, snap-fits) | Free personal | Now has built-in AI text-to-geometry. The right home for the printed part. Charles will be comfortable here. |
| Leo AI | AI copilot in CAD — sketch/spec → assemblies | Paid | Goes beyond single parts to multi-part assemblies (lid + base + button). |
| Zoo (zoo.dev) | Text-to-CAD → real solid geometry | Free / credits | "A 60×40×18mm box with a 12mm button hole…" → CAD that exports to slicers. |
| Blueprint.am | System spec / BOM / wiring (not enclosure CAD) | Free tier tested | Generates parts list, electrical + mechanical connections, build-guide skeleton. Slots in before CAD. See assessment below. |
Prices indicative, Jun 2026. Sources: RapidDirect, Leo AI, Meshy, TRELLIS, Vizcom reviews, The CAD Hub.
Free-tier run generated a real spec pack (BOM, electrical/mechanical connections, build-guide skeleton) for a "safety keychain alert." It is a system/BOM generator, not an enclosure-CAD tool — useful, but read critically.
Battery in the generated spec: 3.7V × 300mAh ≈ 1.1Wh, usable ~240mAh. Rough draws on this design:
| Function | Draw (from the cell) | Continuous runtime on 240mAh |
|---|---|---|
| Siren (120dB, 12V via boost) | ~400mA – 1A | ~15–35 min ⚠ |
| BLE (ESP32-C3, active) | ~40–80mA | several hrs |
| Red SOS flash (5mm, ~20% duty) | ~3–6mA avg | days |
| White torch LED (in-hand readiness) | ~20mA (5mm) · ~100–300mA (B500-style bright) | 12 hrs → ~1–2 hrs if bright |
The torch's value isn't illumination — it's readiness. If the device is already in your hand (because you're using the light walking home), the SOS is instant. The alternative — scrambling for your phone, unlocking it, finding the app — is slower and makes the phone itself the thing worth stealing. The torch is the everyday reason the device is out and primed: it's the physical expression of Phase 1 "On Guard."
On power: a single 5mm LED (~20mA) barely registers; the siren is the dominant load by 30–50×. A real B500-style bright torch (3 LEDs, 100–300mA) does draw meaningfully — so it's managed, not dropped:
The safety-critical trio always comes first on the power budget: Siren Red SOS flash Bluetooth — the torch sits below them, behind the reserve. Then:
| Action | Why |
|---|---|
| Bump the battery to 500–1000mAh | 300mAh is undersized for a 120dB siren. A bigger cell still fits a keychain fob and gets ~45–90 min of siren — the difference between a toy and a credible safety device. |
| Consider a 3–5V piezo (drop the boost) | Removes the MT3608 boost converter entirely — one fewer part, less loss, simpler board. |
| Firmware power policy | Deep-sleep the MCU until triggered (idle → µA, lasts months between charges). Low-battery lockout that reserves ~30% capacity exclusively for siren + SOS + BLE, so non-critical functions can never strand you. |
| Pair a small USB-C charger / keyring power bank | The device already has USB-C — so a tiny pocket charger lets you top up anywhere and removes the residual "what if it's flat" worry (the one risk daily torch use raises). Also a natural bundle add-on that lifts average order value. |
| Step | Detail |
|---|---|
| 1 · Measure first | Exact dimensions of every real part — PCB, button, LED, battery, USB-C. The enclosure is built around these numbers. (Charles to spec the electronics envelope.) |
| 2 · Concept | Sketch 3–4 shape directions in Vizcom / Krea, render them. This is the look & feel to share with Aaron. |
| 3 · Agree the form | Pick a direction with Aaron; optionally one Meshy 3D mockup he can spin to confirm the shape. |
| 4 · Build the real enclosure | Fusion 360 (+ Zoo / Leo for a fast parametric block) to the measured components → export STL/3MF → print on Charles's Ultimaker. |
| 5 · Iterate fit | Cheap test prints to dial in the component fit before any final material. |