A single-file PowerShell + WinForms app that unlocks the laptop's built-in infrared (Windows Hello) camera and turns two overlooked integrated sensors — the ambient light sensor and the microphone array — into live, on-screen instruments, including a small acoustic direction "radar."
No install, no dependencies to build: one .ps1 file and a .cmd launcher.
Recording is the only optional extra (uses ffmpeg).
Why this exists: apps like OBS / VLC cannot show the IR camera (it appears black with no formats) because the IR sensor is exposed only through Windows Media Foundation, not DirectShow. This app reads it the correct way.
IR_demo_web.mp4
Full-quality clip: IR_Sound_Light_Tool.mp4 (committed in the repo).
The demo shows, in order:
| # | What you see | Why it's interesting |
|---|---|---|
| 1 | A person appears in near-total darkness on the IR feed | The IR illuminator + sensor see without visible light |
| 2 | Veins on the forearm become visible | Near-IR (~850–940 nm) penetrates a few mm of skin; blood absorbs NIR more than surrounding tissue, so veins show as dark lines. A normal webcam has an IR-cut filter and cannot show this (same principle as medical vein-finders) |
| 3 | Shining a phone flashlight at the light sensor makes the lux value jump | Live read of the ambient light sensor |
| 4 | Rubbing two hands together on the left, then the right | The radar needle swings LEFT/RIGHT to follow the sound and the dB meter rises with the rubbing |
- Infrared camera viewer (Media Foundation) — works where DirectShow/OBS can't.
- De-flicker — the IR emitter strobes for Windows Hello; the app drops the dark frames so you get a steady image instead of flashing.
- Auto-brighten + Brightness / Contrast / Gamma sliders + digital zoom + mirror + fullscreen.
- Snapshot (PNG) and video recording (MP4 via ffmpeg) — saved to the Desktop.
- Light meter — live ambient light in lux.
- Sound meter — live level (dB) + a radar gauge estimating the LEFT / RIGHT direction of a sound, using the microphone array.
| Laptop | Lenovo ThinkBook 16 G7+ AKP (type 21TJ) |
| CPU | AMD Ryzen AI 7 H 350 w/ Radeon 860M (8 cores / 16 threads) |
| GPU | AMD Radeon 860M |
| RAM | 32 GB |
| OS | Windows 11 Home, build 26200 |
| IR camera | Chicony Integrated IR Camera (VID_04F2&PID_B829), driver Realtek DMFT – IR — 640×360 @30fps, 8-bit grayscale (L8) |
| Light sensor | Ambient Light Sensor via AMD Sensor Fusion Hub |
| Microphone | Realtek Microphone Array (2 channels, 48 kHz) |
The app is generic, but each feature depends on hardware you actually have:
- Windows 10 or 11 with Windows PowerShell 5.1 (built in — no install).
- IR camera viewing: an integrated IR / Windows Hello camera that Windows
exposes through Media Foundation (
SourceKind = Infrared). Most Windows-Hello laptops qualify. Resolution/format are whatever your sensor offers. - Light meter (optional): an ambient light sensor (WinRT
LightSensor). - Sound meter + direction (optional): a 2-channel microphone array. The direction estimate is calibrated per-machine and only resolves a LEFT/RIGHT axis (see How it works).
- Recording (optional): ffmpeg on PATH —
winget install Gyan.FFmpeg.
If a given sensor is missing, that button simply reports it and the rest still works.
1. Download IR-Camera-Viewer.ps1 and IR-Camera-Viewer.cmd (keep them together).
2. Double-click IR-Camera-Viewer.cmd
(it runs: powershell -STA -ExecutionPolicy Bypass -File IR-Camera-Viewer.ps1)
- If the IR view is black and it says "We need to reboot" (
0xC00D7167): restart Windows (this happens after an IR-camera driver change). - Do not replace the IR camera driver with a generic "USB camera" driver — it breaks the Realtek DMFT / Windows Hello path this app relies on.
| Key | Action | Key | Action | |
|---|---|---|---|---|
P |
Snapshot (PNG) | M |
Mirror | |
R |
Record / stop (MP4) | F |
Anti-flicker on/off | |
F11 |
Fullscreen | S |
Auto-brighten on/off | |
+ / - |
Zoom in / out | H |
Show/hide bottom bar | |
Esc |
Exit |
Everything is also on the top menu bar (Functions / Image / Sensors / Help). Sensors are opt-in via the Light and Sound buttons on the top bar; turning Sound off releases the microphone.
- The IR feed appears automatically. If it looks dark, keep Auto-brighten (
S) on, or drag the Bright / Contrast / Gamma sliders on the bottom bar. - Anti-flicker (
F, on by default) hides the IR emitter's strobing so the image is steady instead of flashing. - Mirror (
M) flips left/right, Zoom (+/-) is a digital centre zoom up to 5x, Fullscreen isF11(Escto leave).
- Click Snap (or press
P). Saved to the Desktop asIR_snapshot_<date>_<time>.png— exactly what's on screen (after brightness/contrast/gamma, zoom and mirror).
- Click REC (or press
R) to start — the button turns red and reads STOP. - Click again (or
R) to stop. Saved to the Desktop asIR_rec_<date>_<time>.mp4(H.264, ~15 fps, grayscale — the de-flickered, adjusted image). Needs ffmpeg.
- Light -> live lux; shine a phone flashlight at the sensor to watch it jump.
- Sound -> opens the mic (RAW) and shows the dB level plus the radar needle. Make broadband sound (music, clapping, rubbing hands) on the left/right to see the needle swing. Sensors -> Recenter direction re-zeros the needle when the room is quiet.
- IR = Media Foundation only. The IR sensor reports as
(none)to DirectShow, so OBS/VLC/ffmpeg-dshow see black with no formats. The app uses WinRTMediaFrameReaderwithSourceKind = Infrared. - De-flicker. The IR illuminator strobes on alternate frames (active illumination for face recognition). The app measures per-frame brightness and drops the dark frames, so the effective rate is ~half but the image is steady.
- Image pipeline (C#, compiled at runtime for speed): per-frame auto-contrast stretch + a 256-entry brightness/contrast/gamma lookup table.
- Recording: processed BGRA frames are piped to
ffmpeg(rawvideo → libx264). - Microphone (WASAPI via COM interop): targets the "Microphone Array"
endpoint by name (the default capture device is often the empty 3.5 mm jack).
It opens in RAW mode (
AUDCLNT_STREAMOPTIONS_RAW) to bypass Realtek's noise-suppression/AEC — otherwise quiet ambient sound is gated to exact zero and the two channels become identical. - Sound direction (studied carefully):
- Level difference is not usable here — the array has a large fixed L/R gain imbalance and almost no directional level signal.
- Direction = TDOA (cross-correlation time-delay between the two channels),
auto-centered, mapped
dir = (center − lag) / 3.5. - Use broadband sound (music, claps, rubbing hands, voice). A pure tone does not work: its cross-correlation is periodic (aliased), and the array's per-frequency phase response even flips the apparent direction between frequencies. Broadband averages those phase errors out to the true geometric delay. Measured ground truth (music): LEFT → lag +4, RIGHT → lag −3.
- Two closely-spaced mics ⇒ LEFT/RIGHT axis only, angle approximate — not a full 360° radar.
- IR resolution is fixed by the sensor (here 640×360 grayscale) — it cannot be increased in software.
- IR needs some infrared light; the built-in emitter is short-range (~0.5–1 m). Daylight or an external 850 nm illuminator helps.
- Sound direction is LEFT/RIGHT only and works best with broadband sound.
- A "proper" angle upgrade would need FFT-based GCC-PHAT with a per-frequency phase-offset calibration curve.
Desktop\IR_snapshot_YYYYMMDD_HHMMSS.png and Desktop\IR_rec_YYYYMMDD_HHMMSS.mp4.
The app accesses your camera and (when you enable Sound) your microphone, locally only — nothing is uploaded. Snapshots/recordings stay on your Desktop.
Julian — dohoanghuan@gmail.com
Issues, questions, and pull requests are welcome.
MIT © 2026 Julian — see LICENSE. Provided as-is; hardware IDs and
calibration numbers above are specific to the tested laptop and may differ on yours.