◎ Radiance Color Space Convert

Every color space converts to every other color space, so eventually some pack ships the one node that just does all of them. This is Radiance's. It's a GPU-accelerated matrix-based converter covering the dozen spaces that actually show up in film and VFX work - sRGB, linear sRGB, ACEScg, ACEScct, ACES2065-1, Rec.709, Rec.2020, DCI-P3, Display P3, DaVinci Wide Gamut, ARRI Wide Gamut 4, and S-Gamut3.Cine - and lets you hop between any two.

If that list looks like a roster of "every camera and deliverable I've ever heard of," that's the point. It's the node you reach for when a plate from one pipeline needs to live in another, or when you want to grade in ACEScg and deliver in Rec.709 without running a gauntlet of one-way converters.

How it works

The source builds the conversion as a primaries → XYZ → primaries chain: it computes the RGB-to-XYZ matrix from the source primaries and white point, and the inverse from the target, then multiplies through. The genuinely interesting part is the optional chromatic adaptation - when source and target white points differ, it re-aims the white via Bradford (the default), Von Kries, or plain XYZ scaling. Bradford is what the industry settled on decades ago; leave it alone until you have a reason not to.

Then there's the transfer-function layer: ACEScct and the log-ish spaces get their curves applied separately (the code has explicit ACEScg ↔ ACEScct encode/decode), because a 3×3 matrix can't turn a linear value into a log value.

The inputs that matter

• source_space / target_space - the big dropdown pair. Same 12-space list on both. The only input you really have to set.
• exposure (stops, ±10) and gamma_adjust (0.1–5.0) - applied as part of the conversion, handy for matching a plate's brightness while you're in there. Keep them at 0/1.0 until you actually need them; it's easy to forget a +2 EV you dialed in for testing.
• chromatic_adaptation - Bradford / Von Kries / XYZ Scaling / None. Bradford default; change it only if you know what the difference is.
• use_gpu (default true) - the conversion can run as a matrix on the GPU instead of a numpy loop.

Output is a single image, same shape in, same shape out.

The trap to avoid

Note what this node doesn't do: it converts between linear spaces and log-ish working spaces like ACEScct, but it is not a display transform and it won't tone-map HDR down to SDR for you. If you take an HDR float image and convert it to sRGB here, the output is still the same dynamic range - just re-based - and it'll look crushed or blown until you run it through Radiance's tone map or the ACES 2.0 output transform. In other words: this is the middle of your chain, not the end.

Also worth knowing: S-Gamut3.Cine, AWG4 and DaVinci Wide Gamut all being in one list is a hint that Radiance is aimed at people who have camera plates and EXR files lying around, which is exactly the crowd the pack's community thread describes - HDR/EXR work straight in ComfyUI.

Install

ComfyUI Manager → search Radiance, or:

cd ComfyUI/custom_nodes
git clone https://github.com/fxtdstudios/radiance.git
cd radiance
pip install -r requirements.txt   # Linux: install libopenexr-dev first

Restart ComfyUI and it's in the FXTD Studios/Radiance/Color menu.