diff --git a/include/RaZ/Math/Vector.hpp b/include/RaZ/Math/Vector.hpp
index dc74319c..e03bed1a 100644
--- a/include/RaZ/Math/Vector.hpp
+++ b/include/RaZ/Math/Vector.hpp
@@ -89,10 +89,10 @@ class Vector {
   /// \tparam SqLengthT Type of the squared length value. For vectors of an integral type, it is defined to a large unsigned integer type to avoid overflows.
   /// \return Vector's squared length.
   template <typename SqLengthT = std::conditional_t<std::is_integral_v<T>, uint64_t, T>>
-  constexpr SqLengthT computeSquaredLength() const noexcept { return dot(*this); }
+  constexpr SqLengthT computeSquaredLength() const noexcept { return static_cast<SqLengthT>(dot(*this)); }
   /// Computes the length of the vector.
   /// Calculating the actual length requires a square root operation to be involved, which is expensive.
-  /// As such, this function should be used if actual length is needed; otherwise, prefer computeSquaredLength().
+  /// As such, this function should be used only if the actual length is needed; otherwise, prefer computeSquaredLength().
   /// \tparam LengthT Type of the length value. For vectors of an integral type, it is defined to float; otherwise, it is the same as the original vector's.
   /// \return Vector's length.
   template <typename LengthT = std::conditional_t<std::is_integral_v<T>, float, T>>
diff --git a/include/RaZ/Math/Vector.inl b/include/RaZ/Math/Vector.inl
index 00d0204e..26c17df9 100644
--- a/include/RaZ/Math/Vector.inl
+++ b/include/RaZ/Math/Vector.inl
@@ -103,8 +103,12 @@ template <typename NormedT>
 constexpr Vector<NormedT, Size> Vector<T, Size>::normalize() const noexcept {
   static_assert(std::is_floating_point_v<NormedT>, "Error: The normalized vector's type must be floating-point.");
 
+  const NormedT sqLength = computeSquaredLength<NormedT>();
   Vector<NormedT, Size> res(*this);
-  res /= computeLength();
+
+  if (sqLength != static_cast<NormedT>(0))
+    res /= std::sqrt(sqLength);
+
   return res;
 }
 
diff --git a/tests/src/RaZ/Math/Vector.cpp b/tests/src/RaZ/Math/Vector.cpp
index 4137d007..3ccfc6de 100644
--- a/tests/src/RaZ/Math/Vector.cpp
+++ b/tests/src/RaZ/Math/Vector.cpp
@@ -265,6 +265,12 @@ TEST_CASE("Vector normalization") {
 
   CHECK(std::is_same_v<decltype(std::declval<Raz::Vec3d>().normalize()), Raz::Vec3d>);
   CHECK_THAT(vec3d1.normalize(), IsNearlyEqualToVector(Raz::Vec3d(-0.0037504754254997, 0.9846554128425506, -0.174469630291203)));
+
+  // Normalizing a 0 vector returns it as is
+  CHECK(Raz::Vec3b(0).normalize() == Raz::Vec3f(0.f));
+  CHECK(Raz::Vec3i(0).normalize() == Raz::Vec3f(0.f));
+  CHECK(Raz::Vec3f(0.f).normalize() == Raz::Vec3f(0.f));
+  CHECK(Raz::Vec3d(0.0).normalize() == Raz::Vec3d(0.f));
 }
 
 TEST_CASE("Vector interpolation") {
diff --git a/tests/src/RaZ/Render/Camera.cpp b/tests/src/RaZ/Render/Camera.cpp
index a35462ca..ad5f2b79 100644
--- a/tests/src/RaZ/Render/Camera.cpp
+++ b/tests/src/RaZ/Render/Camera.cpp
@@ -32,15 +32,11 @@ TEST_CASE("Camera look-at") {
 
   const Raz::Mat4f& viewMat = camera.computeLookAt(Raz::Vec3f(0.f));
 
-  // If target == position, creates a look-at matrix filled with NaNs:
-  // [ nan, nan, nan, nan ]
-  // [ nan, nan, nan, nan ]
-  // [ nan, nan, nan, nan ]
-  // [ 0,   0,   0,   1   ]
-  for (std::size_t widthIndex = 0; widthIndex < 4; ++widthIndex) {
-    for (std::size_t heightIndex = 0; heightIndex < 3; ++heightIndex)
-      CHECK(std::isnan(viewMat.getElement(widthIndex, heightIndex)));
-  }
+  // If target == position, creates a look-at matrix filled with 0s except for the last row
+  CHECK(viewMat.strictlyEquals(Raz::Mat4f(0.f, 0.f, 0.f, 0.f,
+                                          0.f, 0.f, 0.f, 0.f,
+                                          0.f, 0.f, 0.f, 0.f,
+                                          0.f, 0.f, 0.f, 1.f)));
 
   camera.setTarget(Raz::Vec3f(0.f, 0.f, -1.f));
   camera.computeLookAt(Raz::Vec3f(0.f));