The idea

“Music-Reactive Lighting Startup — Master Summary Updated April 2026 What We're Building An AI/ML-powered lighting automation system that synchronizes lights with live music in real-time. The goal is not simple beat-reactive automation, but lighting that approaches the quality of a professional lighting designer (LD) — understanding musical structure, timbre, emotion, and narrative arc — at a fraction of the cost. Product Architecture Hardware Proprietary non-DMX lights (simpler entry point; DMX adapter as optional add-on) Low-cost communication device bridging lights and server Physical enclosure/box (design TBD — potential in-house capability) Software LayerDescriptionAudio analysis engineReal-time ML processing: tempo, BPM, amplitude, frequency, onset, timbre, structure, articulation, harmonyLighting engineTranslates audio features into lighting commands via shadersSpatial mappingReads venue layout to optimize fixture placement and effectsVisualizerSoftware simulation of the lighting system; public-facing feature AND primary supervised learning environmentMobile appiOS/Android control interfaceWebsiteMarketing, onboarding, documentation ML Strategy Feature Extraction Roadmap PhaseFeaturesMVP (now)Tempo, beat, dynamics, frequency content6 monthsTimbre, articulation, structure, harmony12+ monthsEmotional valence, genre-specific semantics, narrative arc Training Data Methodology A daily pairwise comparison workflow using the visualizer as the data collection environment: 3 shader pairs per session (6 shaders), split-screen, same music playing to both 4 songs per session → ~12 comparisons, ~12 minutes/day 4–6 raters daily, cycling through a preset venue layout library Elo/tournament structure: high-performing shaders re-enter the pool; surfaces which shader metric combinations resonate best Controls: randomize shader left/right position, randomize from 15–20 track playlist across sub-genres, minimum 8–10 exposures before a shader is re-looped Track win rates per genre/song type, not just overall Key Unresolved Items Before Data Collection Begins Define rating metrics (what raters are explicitly scoring) Define shader parameter axes to vary systematically (color, speed, intensity, etc.) External rater recruitment plan (Montreal scene participants) Statistical pipeline: how Bradley-Terry scores feed back into shader pool selection Academic Grounding FieldRelevanceSensory EvaluationPanel design, fatigue, inter-rater reliabilityPreference ElicitationPairwise comparison methodologyPsychometricsRating instrument designBradley-Terry ModelStatistical backbone for tournament rankingPsychophysicsHuman perception of stimuli Top read: Lawless & Heymann — Sensory Evaluation of Food (highest ROI); Thurstone (1927) for theoretical foundation. Team PersonRoleCore SkillsCo-founder #1Business, product, UX/UI, marketing, strategyPhotography, tech consultingCo-founder #2Technical leadHardware, software, ML (depth TBD)Person #3Shader development & aestheticsVideo game artist, pure mathematics background; flex capacity toward physical box design or MLPerson #4Frontend developmentWebsite, eventual mobile apps (iOS/Android)Person (flex)Physical hardware installation & logisticsOn-site installs, transportationPerson (flex)Supervised learning & MLCurrently learning; working on training pipelineInterviewingSocial media specialist—InterviewingAccountant— Key Capability Gaps to Watch Deep audio ML expertise (MILA as potential hiring/advising source) Installation does not scale — temporary moat or bridge strategy needed Hardware engineering capacity (in-house vs. outsource) Business Model StreamDescriptionHardware salesOne-time cost, accessible pricingInstallation serviceNominal fee; specialist maps venue spatiallySaaS subscriptionPrimary recurring revenueRevenue-sharing (under consideration)1–2% of door/bar for DIY venues instead of flat fee Target Market TierWhoWhyPrimarySmall/medium venues, independent DJs, underground ravesCan't afford $500–2000+/event LDSecondaryProfessional LDs wanting AI co-pilotEfficiency, not replacementTertiaryLarge touring artistsCustomization without full-time LD Competitive Landscape CompetitorModelKey NotesSoundSwitch$10–15/month subscriptionIndustry leader, InMusic-owned, major DJ software integrationMaestroDMX~$500–600 one-timeClosest threat: standalone AI hardware, mobile app, no laptop neededAULIOSSubscriptionEuropean, DMX-focused, claims "first AI club lighting"Lightjams, DMXDesktop, A.I. LightshowVariousAutomation tools, mixed pricing Key gaps competitors don't fill: white-glove installation/spatial mapping, Montreal market presence, truly deep music understanding beyond beat-reactive automation. Go-To-Market Beachhead: Montreal underground electronic music scene (raves, warehouse parties, small clubs) Expansion path: Montreal → Toronto → Quebec City → Canada → US border cities Ecosystem: MILA, McGill CIRMMT, Solotech, Ampleman, SAT, MTELUS, Stereo, Igloofest, MUTEK, MEG Montreal”