Three nodes for VFX/grading workflows, all under MaskEditControl/Color.
None of these nodes pull in extra dependencies — math is done in pure
torch using built-in matrices and OETFs.
Converts an IMAGE between four spaces:
| Space | Notes |
|---|---|
srgb |
sRGB-encoded (default ComfyUI image convention). |
linear |
Linear-light (sRGB primaries). |
rec709 |
Rec.709 OETF-encoded. |
acescg |
ACEScg primaries, linear. |
Conversion always routes through linear-sRGB internally. Output is
clamped to [0,1] for display-encoded targets and unclamped for linear
and acescg.
Loads an Adobe .cube LUT file (1D or 3D) and applies it. Trilinear
interpolation for 3D LUTs; per-channel linear interp for 1D. The LUT
parser respects LUT_3D_SIZE, LUT_1D_SIZE, DOMAIN_MIN, DOMAIN_MAX,
and TITLE.
A strength slider blends between the input and the graded result.
Photographic exposure in stops, white-balance temp/tint, and contrast around a configurable mid-grey pivot.
By default operate_in_linear=True, which means the input is treated as
sRGB-encoded, linearized for the math, then re-encoded. Disable this
flag for legacy display-referred grading.