Beyond color: see what the eye sees.
Mix a little light. See how your eye responds. Explore how three kinds of cone cells help us experience a world of color. Rods provide sensitivity to brightness in dim light.
About the model and its limits
Sliders set relative RGB emission and brightness. The preview comes directly from these values in linear light, re-encoded to sRGB. Independently, the emitted spectrum is integrated against each receptor sensitivity and normalized to that receptorโs full-white response. Rod excitation receives 10ร gain and is capped at 100%; each cone excitation x is passed through f(x) = x ร xโต / (0.0001 + xโต). The function uses fractional excitation, not the slider percentage, and does not alter the preview.
RGB does not specify a unique spectrum. This illustration assumes hypothetical pixel spectra and an illustrative low-excitation cone threshold. It does not model perception, adaptation over time, individual differences or your screenโs calibration. Brightness is not an absolute physical unit.
The spectrum is an illustrative sum of Gaussian RGB emission bands centered at 625, 535 and 460 nm (widths ฯ = 18, 22 and 16 nm). Their amplitudes follow linear-light RGB and brightness. It is not a measured spectrum or an exact spectral match to sRGB. Responses integrate this spectrum against the displayed peak-normalized receptor curves.
Beyond color: see what the eye sees.
Choose a pigment. Explore the light it reflects and how that spectrum stimulates the S, M and L cones. Rods provide sensitivity to brightness in dim light.
Follow the calculation
1. Align the wavelengths. Join neighboring samples with straight lines to evaluate all tables from 390 to 780 nm in 1 nm steps.
2. Find reflectance R. Indigo supplies measured-reflectance samples; lighter indigo is five times the original indigo reflectance, with no white component. Absorbance inputs transmit 10โปแดฌ per pass. Two passes through a layer over white backing give R = 10โปยฒแดฌ. For example, A = 1 gives R = 0.01 (1%).
3. Calculate the light reaching the eye: L(ฮป) = E(ฮป) ร R(ฮป), where E is the selected incident spectrum. All sources are scaled to the same โซ E yฬ dฮป (white-reflector luminance). Reflectance describes the material; L describes the outgoing light.
4. Integrate outgoing light against each receptor sensitivity C and divide by that receptorโs white response under the selected lamp at full brightness. Scale by brightness. Rods use min(1, 10x), giving 10ร gain and saturation at 100%; cones use f(x) = x ร xโต / (0.0001 + xโต), with x expressed as a fraction. Multiply by 100 for the bars.
The integral means area under a curve. Numerically, add trapezoids: โซ f dฮป โ ฮฃ (ฮปแตขโโ โ ฮปแตข) ร (fแตข + fแตขโโ) / 2. Each interval here is 1 nm.
5. Calculate the preview separately. Integrate outgoing light against CIE xฬ, yฬ, zฬ and divide by โซ E yฬ dฮป, giving white Y = 1 under the selected lamp. Convert XYZ to linear sRGB for a D65 display, clip to 0โ1, then encode and multiply by 255. No chromatic adaptation is applied: a white reflector may look warm or tinted. The S/R/M/L bars are not RGB channels. Scale linear RGB by brightness before clipping and encoding. Cone thresholds do not affect this preview.
Why can jeans look lighter?
The original indigo curve is already a dark dyed-cotton sample, not pure indigo dye. Its measured reflectance is only about 2.5โ8%. Denim often combines indigo-dyed warp yarns with undyed filling yarns. Indigo ring dyeing leaves pale yarn cores; wear and finishing remove surface color and reveal more of that pale cotton. Dye loading and fabric construction also vary, so one dark sample cannot represent every pair of jeans.
The lighter example uses R_light = 5 ร R_dark, with no added white component. Its reflectance spans 12.5โ40%, so no spectral clipping is needed. Multiplication preserves the spectral shape and chromaticity before display gamut clipping, while making the reflected light five times stronger. This is a brightness-scaling illustration, not a measured jeans spectrum or a model of dye concentration or wear.
Data and model limits
Indigo uses reflectance digitized from a published measurement of indigo-dyed cotton. It represents a dark fabric blue, not pure pigment or every denim wash. Anthocyanin, chlorophyll and carotenoids use published absorption shapes to model a non-scattering layer over a perfect white backing: light crosses the layer twice, giving R = 10โปยฒแดฌ. One-way peak absorbance is set to 1 as a chosen layer strength. This is not measured reflectance of pure molecules or a complete leaf or paint model. Concentration, solvent, pH, substrate and scattering affect real colors; the square is an sRGB approximation.
Anthocyanin, chlorophyll and carotenoids use the solid PROSPECT-D curves in Fรฉret et al. (2017), Figure 8, manually digitized from 400 to 750 nm at 10 nm intervals with extra points at sharp features. Dashed curves and shaded bands are excluded. The 400 nm endpoints are held to 390 nm; the zero 750 nm endpoints are held to 780 nm. These tails are assumptions. Each absorption shape is scaled to peak one-way absorbance 1. Indigo uses the sample No. 1 cotton reflectance in Kawahito & Yasukawa (2009), Figure 3a, approximately digitized every 10 nm from 400 to 700 nm, with constant endpoint extensions to 390 and 780 nm and no peak scaling. Samples are linearly interpolated to 1 nm. CIE and D65 tables are from Colour Science v0.4.6.