M.I.R.A.G.E.

Multi-Input Reconnaissance and Guidance Environment.
A heads-up display you can actually wear.

Overview

I built MIRAGE as a dual- or split-display AR heads-up display for cosplay helmets — though it's portable to plenty of other uses. It renders real-time HUD elements — gauges, compass, pitch ladder, maps, sprites, and live text — onto small displays with passthrough camera video for see-through operation.

The system runs on NVIDIA Jetson or Raspberry Pi, using SDL2 for hardware-accelerated rendering. All HUD elements are defined in plain-text configuration files with hot-reload — no code changes needed to redesign the entire display layout.

Display

HUD Elements

Over 25 HUD element types render real-time data from sensors, GPS, system metrics, and AI state directly onto the display.

  • Analog and digital gauges with animated needles
  • Compass rose with bearing indicator
  • Pitch ladder with horizon line
  • Mini-map with GPS position
  • Animated sprites and transitions
  • Dynamic text with sensor data binding
  • Battery level, CPU/GPU temp, network status
MIRAGE HUD render — gauges, compass, radar, and AI vision overlay (stereoscopic view)
Screens

Up to 16 HUD Layouts

Switch between up to 16 possible HUD layouts, all user-configurable, with animated transitions. Each screen is a configuration file that defines element positions, data bindings, and visual style.

  • Voice-switchable — ask DAWN to change the HUD, hands-free
  • Combat, navigation, diagnostics, and custom layouts
  • Animated transitions between screens
  • Hot-reload — edit config, see changes instantly
  • Per-screen color themes and opacity
MIRAGE HUD — primary flight layout MIRAGE HUD — armor scan layout MIRAGE HUD — environment layout
Capture

Recording & Live Streaming

H.264 encoding captures the HUD output in real time — hardware-accelerated on Jetson, or software encoding on other hosts. Record locally, stream to YouTube Live, or do both at once; with hardware encoding it runs at full framerate with minimal CPU overhead.

  • Hardware (NVENC / V4L2) or software (x264) H.264 encoding
  • Simultaneous record + stream
  • YouTube Live RTMP streaming
  • Configurable bitrate and resolution
First-person view through the helmet — the MIRAGE HUD composited over a real street scene
AI

AI Vision Integration

MIRAGE can capture snapshots from the helmet camera and send them to DAWN for LLM-powered visual analysis. Ask your AI assistant what it sees — from the inside of the helmet.

  • Snapshot capture from helmet camera feed
  • Forwarded to DAWN's vision API for analysis
  • Results displayed on HUD or spoken via TTS
MIRAGE HUD with DAWN's AI vision analysis describing the scene (stereoscopic view)

Hardware

Small displays with passthrough camera video, driven by Jetson or Raspberry Pi.

Component Recommendation Notes
Display Horizon HUD displays (dual) Passthrough camera video, one per eye
Compute Jetson Orin Nano / RPi 5 Jetson for recording + streaming
Camera CSI or USB camera module For AI vision snapshots
View MIRAGE on GitHub