X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;ds=sidebyside;f=overlay_effect.h;h=4c6d37a4fea6d9a95715031426eb7332fcf73346;hb=7af4d1b54ba141fdb74cd13ddc6110708855d157;hp=06ff7939128257e9ccc006e3076b3e91b29aaeb5;hpb=5e05605a85a15944b95d7000745446756cdcbfa8;p=movit

diff --git a/overlay_effect.h b/overlay_effect.h
index 06ff793..4c6d37a 100644
--- a/overlay_effect.h
+++ b/overlay_effect.h
@@ -5,7 +5,7 @@
 // (If both images are the same aspect and the top image has alpha=1.0
 // for all pixels, you will not see anything of the one on the bottom.)
 //
-// This is the “atop” operation from Porter-Duff blending, also used
+// This is the “over” operation from Porter-Duff blending, also used
 // when merging layers in e.g. GIMP or Photoshop.
 //
 // The first input is the bottom, and the second is the top.
@@ -20,6 +20,13 @@ public:
 
 	virtual bool needs_srgb_primaries() const { return false; }
 	virtual unsigned num_inputs() const { return 2; }
+
+	// Actually, if _either_ image has blank alpha, our output will have
+	// blank alpha, too (this only tells the framework that having _both_
+	// images with blank alpha would result in blank alpha).
+	// However, understanding that would require changes
+	// to EffectChain, so postpone that optimization for later.
+	virtual AlphaHandling alpha_handling() const { return INPUT_PREMULTIPLIED_ALPHA_KEEP_BLANK; }
 };
 
 #endif // !defined(_OVERLAY_EFFECT_H)