The technology
It’s not a filter. It’s a simulation.
Every Silver stock is built the way film actually works: light exposes silver-halide crystals, chemistry develops them into dye, and grain is a physical consequence of that process — not a texture laid on top afterward.
The pipeline
Six stages, in the order film actually runs them.
This is the exact order Silver renders in — not a marketing simplification. Getting the order right matters: halation is an optical effect that happens in linear light, before anything compresses tone, and grain is a property of the developed image, so it has to come after chemistry, not before.
01
Linearize
The image is converted into linear light, the space real photons actually add up in, before anything else touches it.
02
Halation
Bright highlights bloom back through the film base and bias warm.
Why this order: halation is optical, and optics happens in linear light — doing it after tone compression gets the physics backwards.
03
Tonemap
A gentle filmic curve reins in whatever halation just pushed past white, so the next stage receives a well-behaved image.
04
Film Chemistry
Spectral crosstalk between layers, a characteristic curve per color layer, dye yield, interlayer coupler development, and base masking: five chemical effects, stacked in the order a real negative develops them.
05
Grain
A field of silver-halide crystals, sized to the stock’s actual ISO, resolved against your image’s own live exposure, not a noise texture.
06
Output
Color-managed back out at full resolution on export — the exact same graph that drove your live preview, run once at native size.
Deep dive
Two things that make it real.
Grain that behaves like grain
Most “film grain” effects are noise: a texture generator makes something grain-shaped and lays it on at a fixed strength. Real grain doesn’t work that way. Its visibility depends on how exposed that patch of the image actually is, and its size depends on the emulsion’s own crystal structure. Silver models both, live.
A stock rated ISO 3200 doesn’t just get “more grain” turned up — its crystals are modeled larger and more variable, exactly the trade a real fast emulsion makes for sensitivity. Load a slow ISO 50 stock and the same math gives you correspondingly fine, tight grain, with no separate dial to fake it.
WHY IT MATTERS
Push a stop, and Silver’s grain reads more visible for the same reason a real underexposed negative’s does, not because a slider got nudged, but because the tonal range grain depends on to read at all just got flatter.
Grounded in the Boolean/Poisson grain model described in Newson, Faraj, Delon & Galerne, “Realistic Film Grain Rendering,” IPOL 7 (2017) — Silver’s implementation is written independently against that published research.
Chemistry, not a color-matched LUT
Color film has three light-sensitive layers, one per primary color, and they don’t behave in isolation — light bleeds between them, each layer’s response rolls off in its own toe and shoulder, and the developing chemistry in one layer chemically restrains its neighbors. That interaction is why a real stock has a color signature, not just a color grade.
Silver models five separate effects: spectral crosstalk between layers, a per-layer characteristic curve (toe / gamma / shoulder), dye yield (how richly a layer’s dye forms, independent of color balance), interlayer coupler development (a real chemical effect that reads as local contrast and saturation), and the film base’s own mask tint.
PUSH & PULL, DONE HONESTLY
Nudge exposure index up or down, and Silver doesn’t just raise contrast, it shifts color balance the same direction a real lab would, because the same per-layer curves drive both.
SEE FOR YOURSELF
Scan vs. simulation.
Drag to compare a real developed-film scan against Silver’s render of the same frame under a matching stock.
