diff --git a/GPU/Common/ShaderUniforms.h b/GPU/Common/ShaderUniforms.h
index 96be41af09..fd13956eec 100644
--- a/GPU/Common/ShaderUniforms.h
+++ b/GPU/Common/ShaderUniforms.h
@@ -194,10 +194,9 @@ struct UB_VS_Bones {
 };
 
 static const char *ub_vs_bonesStr =
-R"(	mat3x4 u_bone[8];
+R"(	mat3x4 u_bone0; mat3x4 u_bone1; mat3x4 u_bone2; mat3x4 u_bone3; mat3x4 u_bone4; mat3x4 u_bone5; mat3x4 u_bone6; mat3x4 u_bone7; mat3x4 u_bone8;
 )";
 
-// HLSL code is shared so these names are changed to match those in DX9.
 static const char *cb_vs_bonesStr =
 R"(	float4x3 u_bone[8];
 )";
diff --git a/GPU/GLES/ShaderManagerGLES.cpp b/GPU/GLES/ShaderManagerGLES.cpp
index 9a7d10e7ef..5d5af4f4a8 100644
--- a/GPU/GLES/ShaderManagerGLES.cpp
+++ b/GPU/GLES/ShaderManagerGLES.cpp
@@ -495,7 +495,7 @@ void LinkedShader::UpdateUniforms(u32 vertType, const ShaderID &vsid, bool useBu
 	float bonetemp[16];
 	for (int i = 0; i < numBones; i++) {
 		if (dirty & (DIRTY_BONEMATRIX0 << i)) {
-			ConvertMatrix4x3To4x4(bonetemp, gstate.boneMatrix + 12 * i);
+			ConvertMatrix4x3To4x4Transposed(bonetemp, gstate.boneMatrix + 12 * i);
 			render_->SetUniformM4x4(&u_bone[i], bonetemp);
 		}
 	}
diff --git a/GPU/GLES/VertexShaderGeneratorGLES.cpp b/GPU/GLES/VertexShaderGeneratorGLES.cpp
index c1916601e8..317c54fbd1 100644
--- a/GPU/GLES/VertexShaderGeneratorGLES.cpp
+++ b/GPU/GLES/VertexShaderGeneratorGLES.cpp
@@ -580,25 +580,14 @@ bool GenerateVertexShaderGLSL(const VShaderID &id, char *buffer, const GLSLShade
 				"w2.x", "w2.y", "w2.z", "w2.w",
 			};
 
-#ifdef USE_BONE_ARRAY
-			if (numBoneWeights == 1)
-				WRITE(p, "  mat4 skinMatrix = w1 * u_bone[0]");
-			else
-				WRITE(p, "  mat4 skinMatrix = w1.x * u_bone[0]");
-			for (int i = 1; i < numBoneWeights; i++) {
-				const char *weightAttr = boneWeightAttr[i];
-				// workaround for "cant do .x of scalar" issue
-				if (numBoneWeights == 1 && i == 0) weightAttr = "w1";
-				if (numBoneWeights == 5 && i == 4) weightAttr = "w2";
-				WRITE(p, " + %s * u_bone[%i]", weightAttr, i);
-			}
-#else
+			const char *boneMatrix = compat.forceMatrix4x4 ? "mat4" : "mat3x4";
+
 			// Uncomment this to screw up bone shaders to check the vertex shader software fallback
 			// WRITE(p, "THIS SHOULD ERROR! #error");
 			if (numBoneWeights == 1)
-				WRITE(p, "  mat4 skinMatrix = w1 * u_bone0");
+				WRITE(p, "  %s skinMatrix = w1 * u_bone0", boneMatrix);
 			else
-				WRITE(p, "  mat4 skinMatrix = w1.x * u_bone0");
+				WRITE(p, "  %s skinMatrix = w1.x * u_bone0", boneMatrix);
 			for (int i = 1; i < numBoneWeights; i++) {
 				const char *weightAttr = boneWeightAttr[i];
 				// workaround for "cant do .x of scalar" issue
@@ -606,20 +595,18 @@ bool GenerateVertexShaderGLSL(const VShaderID &id, char *buffer, const GLSLShade
 				if (numBoneWeights == 5 && i == 4) weightAttr = "w2";
 				WRITE(p, " + %s * u_bone%i", weightAttr, i);
 			}
-#endif
 
 			WRITE(p, ";\n");
 
-			// Trying to simplify this results in bugs in LBP...
-			WRITE(p, "  vec3 skinnedpos = (skinMatrix * vec4(position, 1.0)).xyz %s;\n", factor);
-			WRITE(p, "  vec3 worldpos = (u_world * vec4(skinnedpos, 1.0)).xyz;\n");
+			WRITE(p, "  vec3 skinnedpos = (vec4(position, 1.0) * skinMatrix).xyz %s;\n", factor);
+			WRITE(p, "  vec3 worldpos = (vec4(skinnedpos, 1.0) * u_world).xyz;\n");
 
 			if (hasNormal) {
-				WRITE(p, "  mediump vec3 skinnednormal = (skinMatrix * vec4(%snormal, 0.0)).xyz %s;\n", flipNormal ? "-" : "", factor);
+				WRITE(p, "  mediump vec3 skinnednormal = (vec4(%snormal, 0.0) * skinMatrix).xyz %s;\n", flipNormal ? "-" : "", factor);
 			} else {
-				WRITE(p, "  mediump vec3 skinnednormal = (skinMatrix * vec4(0.0, 0.0, %s1.0, 0.0)).xyz %s;\n", flipNormal ? "-" : "", factor);
+				WRITE(p, "  mediump vec3 skinnednormal = (vec4(0.0, 0.0, %s1.0, 0.0) * skinMatrix).xyz %s;\n", flipNormal ? "-" : "", factor);
 			}
-			WRITE(p, "  mediump vec3 worldnormal = normalize((u_world * vec4(skinnednormal, 0.0)).xyz);\n");
+			WRITE(p, "  mediump vec3 worldnormal = normalize((vec4(skinnednormal, 0.0) * u_world).xyz);\n");
 		}
 
 		WRITE(p, "  vec4 viewPos = vec4((vec4(worldpos, 1.0) * u_view).xyz, 1.0);\n");
diff --git a/GPU/Vulkan/VertexShaderGeneratorVulkan.cpp b/GPU/Vulkan/VertexShaderGeneratorVulkan.cpp
index 22613b0594..cb295d4f6a 100644
--- a/GPU/Vulkan/VertexShaderGeneratorVulkan.cpp
+++ b/GPU/Vulkan/VertexShaderGeneratorVulkan.cpp
@@ -361,25 +361,24 @@ bool GenerateVertexShaderVulkanGLSL(const VShaderID &id, char *buffer, std::stri
 				"w2.x", "w2.y", "w2.z", "w2.w",
 			};
 
-			WRITE(p, "  mat3x4 skinMatrix = w1.x * u_bone[0];\n");
+			WRITE(p, "  mat3x4 skinMatrix = w1.x * u_bone0");
 			if (numBoneWeights > 1) {
 				for (int i = 1; i < numBoneWeights; i++) {
-					WRITE(p, "    skinMatrix += %s * u_bone[%i];\n", boneWeightAttr[i], i);
+					WRITE(p, " + %s * u_bone%d", boneWeightAttr[i], i);
 				}
 			}
 
 			WRITE(p, ";\n");
 
-			// Trying to simplify this results in bugs in LBP...
-			WRITE(p, "  vec3 skinnedpos = (vec4(position, 1.0) * skinMatrix) %s;\n", factor);
-			WRITE(p, "  vec3 worldpos = vec4(skinnedpos, 1.0) * u_world;\n");
+			WRITE(p, "  vec3 skinnedpos = (vec4(position, 1.0) * skinMatrix).xyz %s;\n", factor);
+			WRITE(p, "  vec3 worldpos = (vec4(skinnedpos, 1.0) * u_world).xyz;\n");
 
 			if (hasNormal) {
-				WRITE(p, "  mediump vec3 skinnednormal = vec4(%snormal, 0.0) * skinMatrix %s;\n", flipNormal ? "-" : "", factor);
+				WRITE(p, "  mediump vec3 skinnednormal = (vec4(%snormal, 0.0) * skinMatrix).xyz %s;\n", flipNormal ? "-" : "", factor);
 			} else {
-				WRITE(p, "  mediump vec3 skinnednormal = vec4(0.0, 0.0, %s1.0, 0.0) * skinMatrix %s;\n", flipNormal ? "-" : "", factor);
+				WRITE(p, "  mediump vec3 skinnednormal = (vec4(0.0, 0.0, %s1.0, 0.0) * skinMatrix).xyz %s;\n", flipNormal ? "-" : "", factor);
 			}
-			WRITE(p, "  mediump vec3 worldnormal = normalize(vec4(skinnednormal, 0.0) * u_world);\n");
+			WRITE(p, "  mediump vec3 worldnormal = normalize((vec4(skinnednormal, 0.0) * u_world).xyz);\n");
 		}
 
 		WRITE(p, "  vec4 viewPos = vec4((vec4(worldpos, 1.0) * u_view).xyz, 1.0);\n");
@@ -408,7 +407,7 @@ bool GenerateVertexShaderVulkanGLSL(const VShaderID &id, char *buffer, std::stri
 		bool distanceNeeded = false;
 
 		if (enableLighting) {
-			WRITE(p, "  vec4 lightSum0 = u_ambient * %s + vec4(u_matemissive, 0.0);\n", ambientStr);
+			WRITE(p, "  lowp vec4 lightSum0 = u_ambient * %s + vec4(u_matemissive, 0.0);\n", ambientStr);
 
 			for (int i = 0; i < 4; i++) {
 				GELightType type = static_cast<GELightType>(id.Bits(VS_BIT_LIGHT0_TYPE + 4 * i, 2));