diff --git a/GPU/Software/Lighting.cpp b/GPU/Software/Lighting.cpp
index e5f8b0fe2c..ff920c17fe 100644
--- a/GPU/Software/Lighting.cpp
+++ b/GPU/Software/Lighting.cpp
@@ -38,7 +38,25 @@ void Process(VertexData& vertex)
 	float maa = (gstate.materialupdate&1) ? gstate.getMaterialAmbientA()/255.f : vertex.color0.a();
 	vertex.color0.a() = gstate.getAmbientA()/255.f * maa;
 
-	// Currently only implementing ambient lighting, so secondary color is always zero anyway
+	for (unsigned int light = 0; light < 4; ++light) {
+		if (!gstate.isLightChanEnabled(light))
+			continue;
+
+		Vec3<float> ldc = Vec3<float>(gstate.getDiffuseColorR(light), gstate.getDiffuseColorG(light), gstate.getDiffuseColorB(light))/255.f;
+		Vec3<float> mdc = (gstate.materialupdate&2)
+							? Vec3<float>(gstate.getMaterialDiffuseR(), gstate.getMaterialDiffuseG(), gstate.getMaterialDiffuseB())/255.f
+							: vertex.color0.rgb();
+		Vec3<float> L = Vec3<float>(getFloat24(gstate.lpos[3*light]&0xFFFFFF), getFloat24(gstate.lpos[3*light+1]&0xFFFFFF),getFloat24(gstate.lpos[3*light+2]&0xFFFFFF));
+		L -= vertex.worldpos;
+
+		float factor = Dot(L,vertex.normal) / L.Length() / vertex.worldpos.Length();
+
+		vertex.color0.r() += ldc.r() * mdc.r() * factor;
+		vertex.color0.g() += ldc.g() * mdc.g() * factor;
+		vertex.color0.b() += ldc.b() * mdc.b() * factor;
+	}
+
+	// Currently only implementing ambient+diffuse lighting, so secondary color is always zero anyway
 	//if (!gstate.isUsingSecondaryColor())
 	{
 		vertex.color1 = Vec3<float>(0.f, 0.f, 0.f);
diff --git a/GPU/Software/TransformUnit.cpp b/GPU/Software/TransformUnit.cpp
index f8d7a1e863..113290c9fa 100644
--- a/GPU/Software/TransformUnit.cpp
+++ b/GPU/Software/TransformUnit.cpp
@@ -115,6 +115,12 @@ void TransformUnit::SubmitPrimitive(void* vertices, void* indices, u32 prim_type
 				data[i].texturecoords = Vec2<float>(uv[0], uv[1]);
 			}
 
+			if (vreader.hasNormal()) {
+				float normal[3];
+				vreader.ReadNrm(normal);
+				data[i].normal = Vec3<float>(normal[0], normal[1], normal[2]);
+			}
+
 			if (vreader.hasColor0()) {
 				float col[4];
 				vreader.ReadColor0(col);
@@ -133,7 +139,8 @@ void TransformUnit::SubmitPrimitive(void* vertices, void* indices, u32 prim_type
 
 			if (!gstate.isModeThrough()) {
 				ModelCoords mcoords(pos[0], pos[1], pos[2]);
-				data[i].clippos = ClipCoords(ClipCoords(TransformUnit::ViewToClip(TransformUnit::WorldToView(TransformUnit::ModelToWorld(mcoords)))));
+				data[i].worldpos = WorldCoords(TransformUnit::ModelToWorld(mcoords));
+				data[i].clippos = ClipCoords(ClipCoords(TransformUnit::ViewToClip(TransformUnit::WorldToView(data[i].worldpos))));
 				data[i].drawpos = DrawingCoords(TransformUnit::ScreenToDrawing(TransformUnit::ClipToScreen(data[i].clippos)));
 
 				Lighting::Process(data[i]);
diff --git a/GPU/Software/TransformUnit.h b/GPU/Software/TransformUnit.h
index 95d9900ade..9e7c8163bb 100644
--- a/GPU/Software/TransformUnit.h
+++ b/GPU/Software/TransformUnit.h
@@ -43,6 +43,8 @@ struct VertexData
 	{
 		#define LINTERP(T, OUT, IN) (OUT) + ((IN - OUT) * T)
 
+		// World coords only needed for lighting, so we don't Lerp those
+
 		clippos.x = LINTERP(t, a.clippos.x, b.clippos.x);
 		clippos.y = LINTERP(t, a.clippos.y, b.clippos.y);
 		clippos.z = LINTERP(t, a.clippos.z, b.clippos.z);
@@ -54,6 +56,10 @@ struct VertexData
 		texturecoords.x = LINTERP(t, a.texturecoords.x, b.texturecoords.x);
 		texturecoords.y = LINTERP(t, a.texturecoords.y, b.texturecoords.y);
 
+		normal.x = LINTERP(t, a.normal.x, b.normal.x);
+		normal.y = LINTERP(t, a.normal.y, b.normal.y);
+		normal.z = LINTERP(t, a.normal.z, b.normal.z);
+
 		color0.x = LINTERP(t, a.color0.x, b.color0.x);
 		color0.y = LINTERP(t, a.color0.y, b.color0.y);
 		color0.z = LINTERP(t, a.color0.z, b.color0.z);
@@ -64,9 +70,11 @@ struct VertexData
 		color1.z = LINTERP(t, a.color1.z, b.color1.z);
 	}
 
+	WorldCoords worldpos; // TODO: Storing this is dumb, should transform the light to clip space instead
 	ClipCoords clippos;
 	DrawingCoords drawpos; // TODO: Shouldn't store this ?
 	Vec2<float> texturecoords;
+	Vec3<float> normal;
 	Vec4<float> color0; // TODO: Should be an int vector!
 	Vec3<float> color1; // TODO: Should be an int vector!
 };