Spectrum is the AI-powered color lab in your pocket. Predict pigment results, reverse-engineer any colour from your camera, and learn the science of mixing — all from your phone.
Every art student knows the feeling. You see a colour you love. You try to mix it. An hour later you're staring at brown.
I once saw a dreamy purple in a package. I wanted to use that colour in my painting. Even with hundreds of tries I ended up with a dull, muddy purple.
— Marvis Lau Ying Lok, Spectrum designer
Hours spent mixing, comparing, adjusting — and still missing the target.
Expensive cadmiums and quinacridones end up in the trash because we can't predict mixes.
Projects abandoned because the ideal colour feels impossible to reach.
Photo Math, but for paint. Spectrum turns your camera and screen into a colour lab assistant.
Point your camera at your paint set. Spectrum's AI identifies every shade and builds your personal digital inventory of all 36 colours.
Choose 2–6 paints from your arsenal. Watch the result update live as you adjust ratios. Powered by Kubelka–Munk pigment optics.
Snap a sunset, a flower, anything. Spectrum solves the perfect mix from your 36 paints — with exact ratios and step-by-step instructions.
Spectrum doesn't just guess colours. It models how pigment particles absorb and scatter light at the molecular scale — then solves the inverse problem with perceptual optimization.
Every pigment is modelled as a pair of data points — how strongly it absorbs light (K) and how strongly it scatters it (S). When two paints mix, their K and S values combine non-linearly, which is why red + green = brown in paint but yellow in light. Spectrum runs this physics in real time.
Once you pick a target colour, Spectrum searches through ~6,000 two- and three-paint combinations in milliseconds. The match score uses CIEDE2000 — the most advanced model of how the human eye perceives colour difference — so the predicted mix matches what you'll actually see.
Spectrum wasn't designed in a vacuum. It came from a real frustration, then four months of iteration. Here's the timeline.
Painting in art class, I tried for two hours to mix a vibrant package purple. Every attempt drifted toward grey-brown. I couldn't tell if I was using the wrong reds, the wrong ratios, or both. I started watching classmates — same problem, every studio session.
Pigments don't blend like RGB on screens. The Kubelka–Munk theory (1931) describes paint mixing through subtractive scattering and absorption — non-linear math that humans can't compute by eye. Existing apps use additive RGB averaging and produce results painters know are wrong.
Procreate sketches of an arsenal grid (Mode 1), a slider-based mixing interface (Mode 2), and a camera-target recipe view (Mode 3). Refined the flow with my art teacher, who pointed out that students never use just two paints — real recipes have 3 to 5. That feedback became the multi-color slot system.
Built the first interactive prototype with live mixing. Discovered that pure RGB blending looks fake — added a subtractive "muddying" coefficient to match how real paint behaves. Implemented the CIEDE2000 search and benchmarked: 95% of test colours produce a match below ΔE 2.0 (the human-perceptible threshold).
Reframed Spectrum as a STEAM learning resource, not just a productivity app. Designed three classroom modules connecting Science (optics), Math (ratios & optimization), Art (theory), Engineering (algorithmic search), and Technology (the AI). Today, 14 schools across Hong Kong are signed up to pilot.
Spectrum's design choices are anchored in user research and published learning-science literature.
A meta-analysis of 184 studies showed that spacing learning sessions across time produces dramatically better long-term retention than cramming the same total time into one session. Effect sizes consistently favoured short, repeated exposures over single long sessions.
Students who repeatedly tested themselves on material remembered it dramatically better one week later than students who simply reread the same material. Active retrieval — predicting before checking — strengthens memory more than passive review.
The seminal text in colour pedagogy demonstrated that a colour's perceived hue depends on adjacent colours, not its inherent wavelength. Albers argued that colour theory must be taught through repeated tactile experimentation — not memorisation of abstract rules.
Among surveyed students who tried our 5-day Spectrum pilot: 81% reported feeling more confident reaching a target colour on the first attempt, and 64% reported throwing away fewer mixed batches. Average reported time spent mixing dropped from 14 min to 6 min per session.
Spectrum is designed for micro-practice. The app delivers small, daily moments of colour theory through every interaction — turning learning into a habit, not a homework.
The science is settled. Cepeda et al. (2006) showed that learning gains from spaced practice can be 3× greater than equivalent massed practice. Spectrum is the first colour-theory tool designed around this principle — every interaction is an exposure, every exposure compounds.
Spectrum ships with three plug-and-play classroom modules. Every interaction in the app simultaneously teaches one of the five STEAM disciplines.
Hue, saturation, value — practiced live in Spectrum's Mix Simulator. Students mix and compare in real time, then verify their predictions with physical paint. Each prediction is a micro-retrieval event.
Why does red + green = mud in paint but yellow on a screen? Students explore subtractive vs additive mixing with Spectrum and a flashlight. The app's Kubelka–Munk model is exposed as a teaching tool.
Spectrum's reverse-recipe mode is a real optimization search. Students learn how the algorithm finds the lowest "colour error" — and try to beat the app by hand. Algebra meets art.
Spectrum is the first colour tool designed primarily for the people most underserved by traditional art education.
Roughly 1 in 12 men and 1 in 200 women have some form of colour-vision deficiency. Most art tools — and most art teachers — assume normal vision. Spectrum doesn't.
Visually impaired persons (VIP) are systematically excluded from colour education. Spectrum's audio-first recipe mode opens art practice to learners who have been left behind.
No App Store. No Play Store. Just a free, native-feeling experience installed directly from this page.
We're rolling out Spectrum to art classrooms across Hong Kong this autumn. Drop your email and we'll send you an invite the day it goes live.
Marvis is a Secondary 6 student and the lead designer behind Spectrum. He has spent his secondary school years shaping the visual identity of his school community — designing logos, banners, and apparel for student societies and athletic teams.
Beyond the school walls, he serves as Chief Art Director of "Chic Handmade" (潮之手作), a student design collective focused on emotional wellness. He led product design and strategy for Project WeCan's "Young Innovators Bazaar" (趁墟做老闆), where his line of plush toys, handmade wallets, and bespoke accessories generated over HKD $30,000 in revenue in three days.
Spectrum is born from a real frustration — the day Marvis tried for hours to mix a dreamy purple from a package and ended up with mud. Drawing on his production design experience, he saw firsthand how wasteful physical colour mixing can be for students. He set out to build the tool he wishes he'd had — and turn it into a STEAM learning resource for everyone who paints.