colorx reference

# Overview

A faithful clone of Nuke's Expression node: four per-channel text fields (r= g= b= a=) hold math that is compiled once per render and evaluated per pixel. It runs as a filter over its source clip and loads in Autodesk Flame (2021+), DaVinci Resolve / Fusion, Nuke, Natron, and any other OpenFX host. The node appears under FORGE / color / colorx (labelled colorx).

The panel has four pages: Channels (the r/g/b/a expressions), Variables (a block of name = formula statements), Constants (the k1..k4 knobs + Reference Colour), and Output (Mix, Clamp, and the Preset pulldown).

# Install

Grab a prebuilt bundle from the latest release (macOS + Linux), or build from source.

1. Unzip → colorx.ofx.bundle.
2. Copy it to the OFX plugin path (root-owned — use sudo): /Library/OFX/Plugins/ on macOS, /usr/OFX/Plugins/ on Linux.
3. macOS only — clear the download quarantine, or the host silently skips it:

   sudo xattr -dr com.apple.quarantine "/Library/OFX/Plugins/colorx.ofx.bundle"copy

4. Restart the host — OFX plugins are scanned at launch.

# The pixel model

An expression sees only the current pixel — its r g b a and its coordinates. That makes it perfect for generators and per-pixel grades, but it cannot blur, sharpen, or sample neighbouring pixels.

• x y are pixel coordinates, origin bottom-left (as in Nuke).
• cx cy are centred and aspect-preserved: 0,0 at the centre, ±1 at the left/right edges — use these for resolution-independent patterns.
• Channels are normalised 0..1. For byte/short clips they are scaled on the way in and written back at the clip's bit depth (clamped); float clips pass through unclamped (HDR-safe).
• An empty channel evaluates to 0; the defaults r g b a are a pass-through.

# Predefined variables

Available to every expression, alongside any names you define.

# Knobs & output

# Operators

Precedence follows C, with one deliberate choice: ^ binds tighter than unary minus, so -2^2 == -4.

# Function library

Matches Nuke's Expression function set. Type to filter.

# Preset gallery

Every preset below is built into the Preset pulldown — picking one fills the channels, the Variables block, and the suggested k1..k4. They're also copy-paste ready. k1..k4 are the live sliders; ref is the Reference Colour.

r = (x+0.5)/width
g = (y+0.5)/height
b = 0
a = 1copy

vars: d = hypot(cx,cy)
r = clamp(1 - d/k1)
g = clamp(1 - d/k1)
b = clamp(1 - d/k1)
a = 1copy

vars: ang = (atan2(cy,cx) + pi) / (2*pi)
r = ang
g = ang
b = ang
a = 1copy

vars: u = x/width
r = 0.5 + 0.5*cos(6.2831853*(u + 0.00))
g = 0.5 + 0.5*cos(6.2831853*(u + 0.33))
b = 0.5 + 0.5*cos(6.2831853*(u + 0.67))
a = 1copy

vars: u = x/width
r = lerp(ref.r, 1, u)
g = lerp(ref.g, 1, u)
b = lerp(ref.b, 1, u)
a = 1copy

vars: c = fmod(floor(x/k1) + floor(y/k1), 2)
r = c
g = c
b = c
a = 1copy

vars: f = x/k1 - floor(x/k1)
r = step(0.5, f)
g = step(0.5, f)
b = step(0.5, f)
a = 1copy

vars: d = hypot(cx,cy)
r = 0.5 + 0.5*sin(d*k1*6.2831853)
g = 0.5 + 0.5*sin(d*k1*6.2831853)
b = 0.5 + 0.5*sin(d*k1*6.2831853)
a = 1copy

vars: ang = atan2(cy,cx); rad = hypot(cx,cy);
      m = step(rad, 0.4 + 0.3*cos(ang*k2))
r = m
g = m
b = m
a = 1copy

vars: v = sin(cx*k1*3) + sin(cy*k1*3 + t*0.1)
      + sin((cx+cy)*k1*2)
      + sin(hypot(cx,cy)*k1*4 - t*0.13)
r = 0.5 + 0.5*sin(v)
g = 0.5 + 0.5*sin(v + 2.0944)
b = 0.5 + 0.5*sin(v + 4.1888)
a = 1copy

vars: n = fBm(cx*k1, cy*k1, t*0.05, 5, 2, 0.5);
      c = clamp(n*0.5 + 0.5)
r = c
g = c
b = c
a = 1copy

vars: tu = turbulence(cx*k1, cy*k1, t*0.05, 5, 2, 0.5);
      m = 0.5 + 0.5*sin((cx + tu*2)*6.2831853*k2)
r = m
g = m
b = m
a = 1copy

vars: n = fBm(cx*k1, cy*k1, t*0.05, 5, 2, 0.5)*0.5 + 0.5
r = 0.5 + 0.5*cos(6.2831853*(n + 0.00))
g = 0.5 + 0.5*cos(6.2831853*(n + 0.33))
b = 0.5 + 0.5*cos(6.2831853*(n + 0.67))
a = 1copy

r = r + (random(x,    y,    t) - 0.5)*k1
g = g + (random(x+11, y,    t) - 0.5)*k1
b = b + (random(x,    y+7,  t) - 0.5)*k1
a = acopy

r = pow(r, 1/k1)
g = pow(g, 1/k1)
b = pow(b, 1/k1)
a = acopy

vars: lum = 0.2126*r + 0.7152*g + 0.0722*b
r = lum + (r - lum)*k1
g = lum + (g - lum)*k1
b = lum + (b - lum)*k1
a = acopy

vars: lum = 0.2126*r + 0.7152*g + 0.0722*b
r = lerp(ref.r, 1, lum)
g = lerp(ref.g, 1, lum)
b = lerp(ref.b, 1, lum)
a = acopy

r = floor(r*k1)/k1
g = floor(g*k1)/k1
b = floor(b*k1)/k1
a = acopy

vars: d = hypot(cx,cy); v = clamp(1 - d*k1)
r = r*v
g = g*v
b = b*v
a = acopy

vars: s = 0.6 + 0.4*sin(y*k1)
r = r*s
g = g*s
b = b*s
a = acopy

Handy constants: 6.2831853 = 2π · 2.0944 = 2π/3 · 4.1888 = 4π/3. The palette presets use Inigo Quilez's cosine palette.

# Hosts & GPU

The same expression runs on whichever back-end the host supports; results match across all of them.