Hot Ones: The Biometric Gauntlet β Real-Time Optical rPPG Capsaicin Lie Detector with AI Sean Evans & Presage SmartSpectra C++ SDK
This is a submission for the Presage SmartSpectra SDK Challenge: https://dev.to/challenges/presage π― Target Category Most Creative or Delightful Use β featuring live on-device desktop rPPG, a 15-Celebrity Hot Ones Biom
This is a submission for the Presage SmartSpectra SDK Challenge: https://dev.to/challenges/presage
π― Target Category
Most Creative or Delightful Use β featuring live on-device desktop rPPG, a 15-Celebrity Hot Ones Biometric Disruption Leaderboard, an SNL Negative Control, a 4K Bare-Skin Benchmark, and a companion Majestic Hoops Post-Game Fatigue Showcase.
π Live Demos & Direct Video Links (Zero-Install)
Live Cloud Run Interactive Broadcast Studio: https://hot-ones-biometric-gauntlet-473903412658.us-central1.run.app
Side-by-Side Before (Raw) & After (Burned-In Biometric HUD) Video Hub: https://hot-ones-biometric-gauntlet-473903412658.us-central1.run.app/payton
75-Second Comedic Tech & Hackathon Promo Video (720p HD MP4): https://hot-ones-biometric-gauntlet-473903412658.us-central1.run.app/clips/payton_kit/smartspectra_hackathon_promo.mp4
Companion 9:16 Vertical Shorts Lab (Majestic Hoops Fatigue + Hot Ones): https://smartspectra-shorts-recruit-473903412658.us-central1.run.app
π² What I Built: "Hot Ones: The Biometric Gauntlet" (AI Sean Evans Edition)
Can you bluff an optical photoplethysmography (rPPG) AI while drinking a bowl of hot red pepper soup spiked with a full fluid ounceβhalf a bottleβof Tabasco?
Hot Ones: The Biometric Gauntlet (AI Sean Evans Edition) is a standalone desktop and cloud broadcast studio built with Electron 33, Node.js 24, @smartspectra/node-sdk (v3.4.0), and Google Gemini 3.8 Flash (@google/genai). It turns your webcam into a real-time Capsaicin Lie Detector hosted by an AI Sean Evans who interrogates you across a 10-round Scoville Ladderβand roasts you with your own live heart rate, breathing waveforms, heart-rate variability (HRV), and 478-point facial micro-expressions the moment you try to "no-sell" the heat.
π§ͺ The Origin Story: Why "1-Oz Tabasco Red Pepper Soup Mode" Solves rPPG Motion Noise
This project started with a real culinary experiment: dumping a full fluid ounce (2 tablespoons / 30 mL / half a standard bottle) of Tabasco into a bowl of steaming red pepper soup.
Reading the official Presage SmartSpectra Model Cards revealed a key physiological insight:
The Solid Wing Problem (15 Seconds of Jaw-Motion Artifact): SmartSpectra's Pulse Rate (12s window) and HRV (60s NN window) model cards explicitly warn that chewing corrupts optical skin reflection on the cheeks and temples. Eating a traditional chicken wing requires ~15 seconds of heavy chewing before the swallow, delaying clean rPPG lock.
The 1-Oz Tabasco Red Pepper Soup Breakthrough (Zero-Chew 3s Fast Lock):
Zero Mastication = Instant Optical Lock: Sipping spicy red pepper soup requires zero chewing. Within 3β4 seconds of the swallow, the masseter and temporalis muscles are completely stationaryβletting SmartSpectra lock onto cardio.pulseRate, cardio.arterialPressureTrace, and breathing.upperTrace immediately.
Thermal + Capsaicin TRPV1 Synergy: Capsaicin lowers the activation threshold of TRPV1 ion channels below body temperature (37Β°C), while steaming broth (60Β°C+) activates those exact same receptorsβtriggering an immediate sympathetic surge (+18 to +26 BPM heart rate spike, rapid vagal HRV drop, and shallow upper-chest panting).
Acetic Acid Vapor Reflex: Hot broth volatilizes Tabasco's vinegar into the sinuses on approach, triggering involuntary DISGUST and SURPRISE spikes in SmartSpectra's 478-point face.expression classifier before the spoon even leaves your lips.
The studio includes three selectable modes:
π² 1-Oz Tabasco Soup Bowl (Origin / Fast-Lock Mode): 10 escalating pours of Tabasco into hot red pepper soup with a 4-second Pour & Sip window straight into the stationary Biometric Staredown.
π Classic 10-Wing Gauntlet: Traditional 10-wing lineup (The Classic 1,800 SHU β Da Bomb 135,600 SHU β The Last Dab Xperience 2,693,000 SHU) with a 15-second Dab & Bite chew gate where rPPG scoring pauses while 478-point facial flinches are tracked.
πΌ Judge File / Sim Mode (sdk.useCustomInput() + 1-Click Capsaicin Simulator): Built so hackathon judges don't have to drink half a bottle of Tabasco to test every waveform, Protobuf metric, and bluff roast.
βοΈ Physics-Aware 4-Phase State Machine
Built around SmartSpectra's documented signal windows (12s Pulse, 30s Breathing, 60s HRV, >35s EDA Proxy):
Phase 0 β Consent Gate & 30s Baseline Calibration: Locks the challenger's personal resting Pulse (BPM), Breathing Rate (BRPM), HRV (rmssd, sdnn, baevsky), and neutral poker face locally on-device.
Phase 1 β Pour & Sip (4s) or Dab & Bite (15s): Gates rPPG scoring during sipping/chewing so motion artifacts never pollute session peaks, while keeping 478-point Face Analysis (face.expression, face.blinking, face.talking) running at full speed.
Phase 2 β Stationary Biometric Staredown (25β30s): Prompts ValidationCode 0 (kOk) framing while AI Sean Evans (Gemini 3.8 Flash + Gemini 3.8 Flash-Lite TTS) asks a tailored interview question and live charts stream cardio.pulseRate, arterialPressureTrace, cardio.hrv, dual breathing.upperTrace vs. lowerTrace, and eda.trace.
Phase 3 β Live Answer & The "No-Sell Lie Detector": Combines spoken transcript with baseline-vs-live biometric deltas. If you claim "Honestly Sean, this barely has a kick" while your pulse jumped +19 BPM and an 84% DISGUST micro-expression leaked, a "π¨ NO-SELL BLUFF DETECTED" siren drops and AI Sean Evans + Conan's unlicensed physician π©Ί Dr. Arroyo (1998 Guadalajara Med) roast your bluff using your exact HUD numbers.
π₯ The 15-Celebrity Hot Ones Biometric Disruption Leaderboard (+ SNL Negative Control)
We ranked 15 iconic Hot Ones episodes from most biometrically disrupted to ice-cold poker faceβand paired them against a Scientific Negative Control (Maya Rudolph playing BeyoncΓ© on SNL eating an un-spiced 0 SHU NBC prop wing) to prove why multimodal optical rPPG beats vision-only emotion AI:
π Conan O'Brien (5-Act Arc): +34 BPM (68 β 102 BPM) Β· HRV -27 ms (48 β 21 ms) Β· Breath +12.2 br/m (14 β 26.2) Β· 96% Disgust/Strain β Tracked from Wing #1 calmness (68 BPM) through Dr. Arroyo's neck exam (83 BPM) to his Da Bomb bottle chug (96 BPM, 81.1% C++ Disgust) and 102 BPM redline screaming "I'M PERFECTLY FINE!"
π₯ Gordon Ramsay: +29 BPM (72 β 101 BPM) Β· HRV -19 ms Β· Breath +13.5 br/m Β· 95% Disgust/Angry β 262 C++ expression packets (61.7% kOk) as he yells "My heart's beating like a drum!" on Da Bomb.
π₯ Key & Peele (Dual 2-Person Lock): +28 BPM (71 β 99 BPM) Β· HRV -18 ms Β· Real C++ breathing.rate lock at 38.7 br/min (+24.2 br/m!) β Simultaneous 2-guest tracking with Sean Evans host-cut filtering.
π₯ Cardi B: +27 BPM (70 β 97 BPM) Β· HRV -19 ms Β· Breath +12.4 br/m Β· 93% Surprise/Fear β 48.0% C++ ValidationCode 0 lock + 184 C++ expression packets.
Lewis Capaldi: +26 BPM (73 β 99 BPM) Β· HRV -18 ms Β· Breath +11.8 br/m Β· 91% Fear/Grimace β Full sympathetic fight-or-flight surge.
Jennifer Lawrence: +25 BPM (70 β 95 BPM) Β· HRV -17 ms Β· Breath +11.0 br/m Β· 92% Fear/Crying β Lacrimal eye-watering & jaw panic tracked by 478-pt WASM.
Pedro Pascal: +24 BPM (70 β 94 BPM) Β· HRV -16 ms Β· Breath +10.6 br/m Β· 89% Strain β Cranial sweat vasodilatory surge.
Shaquille O'Neal: +24 BPM (70 β 94 BPM) Β· HRV -16 ms Β· Breath +9.8 br/m Β· 88% Eye-Wide Surprise ("I apologize, Kansas!").
Billie Eilish: +23 BPM (69 β 92 BPM) Β· HRV -15 ms Β· Breath +10.2 br/m Β· 86% Surprise/Angry.
Idris Elba: +22 BPM (67 β 89 BPM) Β· HRV -15 ms Β· Breath +9.5 br/m Β· 84% Cough/Glare β 47.7% C++ kOk lock (229/480 frames) + 155 C++ expression packets.
Paul Rudd (Smile Bluff): +21 BPM (66 β 87 BPM) Β· HRV -14 ms Β· 86% Happy Smile Bluff β Masks a +21 BPM capsaicin surge behind a wholesome Duchenne smile (28.3% C++ kOk, 88 packets).
Sydney Sweeney: +20 BPM (68 β 88 BPM) Β· HRV -13 ms Β· 82% Smile Bluff β Polite smile until "the cumulative effect" triggers a +20 BPM jump.
DJ Khaled (Quit on Wing #3): +19 BPM (74 β 93 BPM) Β· Real C++ breathing.rate lock at 38.3 br/min (+23.3 br/m!) β Tapped out on Wing #3 (11,600 SHU).
Aubrey Plaza: +16 BPM (68 β 84 BPM) Β· HRV -11 ms Β· 64% Deadpan Leak β Icy stare still leaks +16 BPM.
π§ Lorde (Coolest / Most Composed): +8 BPM (64 β 72 BPM) Β· HRV -4 ms Β· Breath +2.4 br/m Β· 14% Strain β Undisputed Queen of Capsaicin Composure: 66.9% C++ ValidationCode 0 (321/480 frames) & 180 expression packets confirm she barely flinches!
π Negative Control β Maya Rudolph (SNL 0 SHU Prop Wing): +4 BPM (70 β 74 BPM) Β· HRV -3 ms (49 β 46 ms) Β· Breath +0.8 br/m Β· 93% Acted Facial Panic! β Eats an un-spiced NBC prop wing. Fools vision-only emotion AI (93% panic face), while optical rPPG (+4 BPM, 46 ms HRV) proves zero real capsaicin!
π¬ Hardware Benchmark: 4K Bare-Skin Vasodilation vs. Compressed YouTube Studio Makeup
To prove how @smartspectra/node-sdk v3.4.0 performs without 360p YouTube compression, 2-second camera cuts, and heavy TV studio foundation (~58% green-channel attenuation), we ran a controlled 4K Bare-Skin Thermal Vasodilation Benchmark (3840Γ2160 HEVC, 31.18s, 935 frames, padded into a 640Γ640 square ROI with 28px top headroom):
ValidationCode 0 (kOk) Lock Rate: 87.2% (815 / 935 frames) β Code 12 (kExcessiveMotion) fired only during the exact 3.6s window where a hand wiped sweat across the forehead.
Native C++ Protobuf Packet Yield: 1,020 real-time C++ metric packets across all 10 physiological channels, including 15 locked stable: true cardio.pulseRate readings up to 82.8% C++ confidence (tracking recovery from 95.6 BPM down to 79.7 BPM) and 11 breathing.rate readings (10.5 β 17.7 BRPM).
π‘ Developer Field Notes for @smartspectra/node-sdk Builders
Use Square ROI (640Γ640) or Chest-Padded 9:16 Portrait (360Γ640) for Custom Input (sdk.useCustomInput()): Padding with top headroom keeps the forehead out of Code 11 (kFaceTooHigh) and the chest out of Code 7 (kChestNotVisible), lifting kOk lock rate from ~15% to 87%β100%.
Import presage from @smartspectra/node-sdk/messages: decodeMetrics comes from @smartspectra/node-sdk, while the protobuf namespace (presage.smartspectra.Metrics.toObject(decoded)) lives in @smartspectra/node-sdk/messages.
Call preconfigure(cacheDir) Before Starting: Required by cardio.arterialPressureTrace, eda.trace, and micromotion calculators via require('@smartspectra/node-sdk/js/ffi.js').preconfigure(cacheDir).
Call await sdk.stopAsync() at EOF: Flushes all accumulated vitals packets cleanly from the C++ worker thread when streaming pre-recorded files.
Originally published by Dev.to AI. Aggregated on AIWithGhost for educational purposes β full credit and traffic to the original publisher.