SieveColorSpace

Grade Format (.scs — Sieve Color Space)

Overview

A .scs file stores a color grade as a displacement field on the SCS simplex. Unlike .cube files (which are RGB->RGB tables tied to a specific color space), a .scs grade is gamut-independent: it can be captured in one color space and applied in any other.

File structure

JSON with the following keys:

{
  "version": 1,
  "lut_size": 17,
  "delta_pi": [[[ ... ]]],
  "delta_ell": [[[ ... ]]],
  "metadata": {
    "coverage": 85.3,
    "mean_delta_pi": 0.042,
    "mean_delta_ell": 0.031
  }
}

Fields

Field	Type	Description
version	int	Format version (currently 1)
lut_size	int	Grid resolution N (typically 17 or 33)
delta_pi	float[N][N][N][3]	Chromaticity displacement on simplex
delta_ell	float[N][N][N]	Luminance displacement
metadata.coverage	float	Percentage of LUT cells with data (0-100)
metadata.mean_delta_pi	float	Mean absolute chromaticity shift
metadata.mean_delta_ell	float	Mean absolute luminance shift

How it works

The LUT is indexed by the before image’s RGB values. For each grid point (R, G, B):

• delta_pi[R][G][B] = how much to shift the simplex coordinates
• delta_ell[R][G][B] = how much to shift the luminance

At application time:

1. Convert input pixel to SCS: (pi, ell)
2. Look up (delta_pi, delta_ell) by trilinear interpolation in the LUT
3. Apply: pi_new = pi + intensity * delta_pi
4. Normalize pi_new on the simplex
5. Convert back to RGB

Why it’s gamut-independent

The displacement delta_pi lives on the simplex, not in RGB space. The simplex coordinates are the same regardless of which primaries or transfer curve produced the original RGB values. So a grade captured from Rec.709 footage can be applied to S-Log3 footage — the simplex displacement is translated to the target gamut at step 5.

Typical file sizes

LUT size	File size
17^3	~3 MB
33^3	~50 MB

Creating a .scs from a .cube

Use ChromaPlex (tab 3: Convert .cube) with the “Export .scs” checkbox, or programmatically:

import scs

# From before/after images
grade = scs.capture_grade(before_img, after_img, lut_size=17)
scs.save_grade(grade, "my_look.scs")

# From an existing grade
loaded = scs.load_grade("my_look.scs")
result = scs.apply_grade(photo, loaded, intensity=0.8)

Blending two grades

mix = scs.blend_grades(grade_a, grade_b, ratio=0.3)
# 30% of grade A + 70% of grade B (linear on the simplex)
scs.save_grade(mix, "blend_30_70.scs")

This site is open source. Improve this page.