diff --git a/core/dr-film/src/spectrum.rs b/core/dr-film/src/spectrum.rs
index 3a432ea..0ac8a44 100644
--- a/core/dr-film/src/spectrum.rs
+++ b/core/dr-film/src/spectrum.rs
@@ -58,6 +58,18 @@ pub fn illuminant(name: &str) -> Spectrum {
         // the filter's effect is a colour shift, and a colour shift ahead of
         // the balance step is by construction invisible.
         "TH-KG3" | "TH-KG3-L" | "T" => blackbody(3400.0),
+        // A cinema projector's xenon short-arc lamp, which is what a release
+        // print is *looked at* under — Kodak 2383 and 2393 name it.
+        //
+        // Approximated by D55, and named here rather than left to fall through
+        // to the unknown branch, which logged a warning on every bake of a
+        // Vision3 stock and implied something was broken. A xenon arc sits near
+        // 6000 K, close to daylight and nothing like the tungsten above, so
+        // daylight is the right family to stand in for it. It is not a
+        // blackbody — the arc has line structure a Planckian curve has no way
+        // to express — and the residue of that is small here because the
+        // viewing step adapts the white point out either way.
+        "K75P" => ILLUMINANT_D55,
         other => {
             if let Some(kelvin) = other.strip_prefix("BB").and_then(|k| k.parse::<f32>().ok()) {
                 return blackbody(kelvin);
@@ -216,6 +228,21 @@ mod tests {
         }
     }
 
+    #[test]
+    fn every_illuminant_a_shipped_profile_names_is_known() {
+        // The warning that found this said "unknown illuminant K75P, falling
+        // back to D55" on every bake of a Vision3 stock. The fallback was
+        // reasonable and the silence was not: a profile naming a light nobody
+        // implemented should be a failing test rather than a line in a log.
+        for name in ["D50", "D55", "D65", "TH-KG3", "TH-KG3-L", "T", "K75P"] {
+            let rgb = Viewing::new(name).to_srgb(&[1.0; SPECTRUM]);
+            assert!(
+                rgb.iter().all(|c| (c - 1.0).abs() < 0.01),
+                "{name} is not a working illuminant: clear frame gave {rgb:?}"
+            );
+        }
+    }
+
     #[test]
     fn a_hotter_blackbody_is_bluer() {
         // Cheap, but it is the one property that catches Planck's law written