Trip planning still feels like 2012 — too many searches, open tabs, the same route rechecked many times, and packing from memory until you're already at the door.

I travel often. After repeating that loop enough times, when I got Spectacles I kept thinking:

Why is all of this still happening on my phone?

The plan should live in the space around you.
The camera should see what’s in the bag.
The glasses should already know where you’re going.

TripOptic is where this idea landed — a lens that turns a spoken trip brief into a spatial plan you browse with a pinch. No phone. No tab juggling. No context switching.

What’s in the demo

Beat	What you’ll see
Input	Voice-first STT or AR keyboard — stepped fields with same TripDraft
Plan	One Gemini generateContent call → transport, stay, places, food, weather, pack
Browse	SIK pinch → spatial category rows → detail panel (formatted notes & options)
Pack	Camera capture → Gemini Vision grounded in trip + AccuWeather forecast
Destination	RSG Imagen → texture decode → native Spatial Image (layered planes fallback)

Why Spectacles — not a phone, not a chatbot

• Hands-free capture — speak the trip at the airport, mid-pack, or walking out the door
• Spatial plan board — categories live around you instead of inside a vertical scroll
• Contextual pack scan — the camera sees the bag, the plan knows the route, the forecast shapes the suggestions

The interaction changes when the plan exists in your physical space instead of behind glass.

Scope (honest)

Planning and browsing only, not booking (yet)

Transport and stay rows are structured AI suggestions with “check Skyscanner”-style pointers — no live fare APIs in this repo.

API keys are configured via Lens Studio RSG credentials (not committed).

Exploring next: fresh-session search per category tap to reduce repeat lookups on the same route.
This entry focuses on the in-lens: plan → pack validation → destination preview.

Stack

Lens Studio
Spectacles Interaction Kit (SIK)
Remote Service Gateway (Gemini plan + vision, Imagen destination)
AccuWeather
Optional TTS (OpenAI via RSG, local config only)

UI/UX:  AurixEmberwave | ArtStation

GitHub Repo

Roast welcome — especially:

• Weird city/date voice inputs
• Pack scan results on device
• Spatial Image vs layered planes — which feels more native?

If you were building for Spectacles, what would you ship next: Spatial itinerary timeline or flight-day mode?

After about four months since the last public drop, I shipped MiNiMIDI v3 as a proper upgrade — not just new visuals, but problems I’d been turning over in my head: how to keep a BPM slider honest when playback is DynamicAudioOutput + huge Lyria buffers, and how to make the crossfader feel smooth on Spectacles instead of fine in Editor and dead on device. Preview on PC was hiding how mean those paths are to the wearable. There’s also a spectrum ring up top — pinch around it for a theremin-style layer and extra control over the vibe while you’re mixing.

What’s new in this version

• Audio engine (Spectacles): Per-layer level via AudioComponent.volume next to DynamicAudioOutput where possible — avoids full-buffer PCM scaling on every fader move.
• Stability: Serialized DynamicAudioOutput updates (max one heavy pump per frame) + debounced / gated slider paths so crossfader + BPM don’t stack lethal work in one tick.
• Crossfader: SIK-aware (real min/max), owner-resolved layers, lifecycle-safe queues (no delayed pumps after release / replace).
• BPM: Slider path tuned for large stems — debounced resample, commit on release so scrubbing stays usable on device.
• UX / build: Auto-generated pad + control layout from config (far less manual Inspector wiring).
• Expression: Spectrum ring — pinch navigation + theremin-style bed layered on the mix.

Repo: https://github.com/urbanpeppermint/MiniMIDI-v3

Design credit:  https://www.reddit.com/user/AurixEmberwave/ fully redesigned the MIDI console layout