diff --git a/src/python/magnum/math.cpp b/src/python/magnum/math.cpp
index 4b81111..598083b 100644
--- a/src/python/magnum/math.cpp
+++ b/src/python/magnum/math.cpp
@@ -155,6 +155,10 @@ template<class T> void boolVector(py::class_<T>& c) {
     c.def_static("__len__", []() { return int(T::Size); }, lenDocstring);
 }
 
+template<class U, class T, class ...Args> void convertible(py::class_<T, Args...>& c) {
+    c.def(py::init<U>(), "Construct from different underlying type");
+}
+
 template<class T> void quaternion(py::module& m, py::class_<T>& c) {
     /*
         Missing APIs:
@@ -311,15 +315,15 @@ void math(py::module& root, py::module& m) {
 
     /* These are needed for the quaternion, so register them before */
     magnum::mathVectorFloat(root, m);
-    magnum::mathVectorIntegral(root, m);
     magnum::mathMatrixFloat(root);
-    magnum::mathMatrixDouble(root);
 
     /* Quaternion */
     py::class_<Quaternion> quaternion_(root, "Quaternion", "Float quaternion");
     py::class_<Quaterniond> quaterniond(root, "Quaterniond", "Double quaternion");
     quaternion(m, quaternion_);
     quaternion(m, quaterniond);
+    convertible<Quaterniond>(quaternion_);
+    convertible<Quaternion>(quaterniond);
 }
 
 }
diff --git a/src/python/magnum/math.matrix.h b/src/python/magnum/math.matrix.h
index 583b757..d7608cc 100644
--- a/src/python/magnum/math.matrix.h
+++ b/src/python/magnum/math.matrix.h
@@ -154,6 +154,10 @@ template<class T> void matrix(py::class_<T>& c) {
         .def("determinant", &T::determinant, "Determinant");
 }
 
+template<class U, class T, class ...Args> void convertible(py::class_<T, Args...>& c) {
+    c.def(py::init<U>(), "Construct from different underlying type");
+}
+
 template<class T> void matrices(
     py::class_<Math::Matrix2x2<T>>& matrix2x2,
     py::class_<Math::Matrix2x3<T>>& matrix2x3,
diff --git a/src/python/magnum/math.matrixdouble.cpp b/src/python/magnum/math.matrixdouble.cpp
index 9b44cc1..cf8e955 100644
--- a/src/python/magnum/math.matrixdouble.cpp
+++ b/src/python/magnum/math.matrixdouble.cpp
@@ -48,6 +48,21 @@ void mathMatrixDouble(py::module& root) {
         matrix3x2d, matrix3x3d, matrix3x4d,
         matrix4x2d, matrix4x3d, matrix4x4d,
         matrix3d, matrix4d);
+
+    /* At this point we should have both float and double types registered,
+       so register type conversions */
+    convertible<Matrix2x2>(matrix2x2d);
+    convertible<Matrix2x3>(matrix2x3d);
+    convertible<Matrix2x4>(matrix2x4d);
+    convertible<Matrix3x2>(matrix3x2d);
+    convertible<Matrix3x3>(matrix3x3d);
+    convertible<Matrix3x4>(matrix3x4d);
+    convertible<Matrix4x2>(matrix4x2d);
+    convertible<Matrix4x3>(matrix4x3d);
+    convertible<Matrix4x4>(matrix4x4d);
+
+    convertible<Matrix3>(matrix3d);
+    convertible<Matrix4>(matrix4d);
 }
 
 }
diff --git a/src/python/magnum/math.matrixfloat.cpp b/src/python/magnum/math.matrixfloat.cpp
index d0ce98a..8a7069c 100644
--- a/src/python/magnum/math.matrixfloat.cpp
+++ b/src/python/magnum/math.matrixfloat.cpp
@@ -48,6 +48,23 @@ void mathMatrixFloat(py::module& root) {
         matrix3x2, matrix3x3, matrix3x4,
         matrix4x2, matrix4x3, matrix4x4,
         matrix3, matrix4);
+
+    /* Register the double types as well, only after that register type
+       conversions because they need all the types */
+    mathMatrixDouble(root);
+
+    convertible<Matrix2x2d>(matrix2x2);
+    convertible<Matrix2x3d>(matrix2x3);
+    convertible<Matrix2x4d>(matrix2x4);
+    convertible<Matrix3x2d>(matrix3x2);
+    convertible<Matrix3x3d>(matrix3x3);
+    convertible<Matrix3x4d>(matrix3x4);
+    convertible<Matrix4x2d>(matrix4x2);
+    convertible<Matrix4x3d>(matrix4x3);
+    convertible<Matrix4x4d>(matrix4x4);
+
+    convertible<Matrix3d>(matrix3);
+    convertible<Matrix4d>(matrix4);
 }
 
 }
diff --git a/src/python/magnum/math.vector.h b/src/python/magnum/math.vector.h
index 38efe12..4967fdd 100644
--- a/src/python/magnum/math.vector.h
+++ b/src/python/magnum/math.vector.h
@@ -261,6 +261,19 @@ template<class T> void vector4(py::class_<Math::Vector4<T>>& c) {
             "XY part of the vector");
 }
 
+template<class U, template<class> class Type, class T, class ...Args> void convertibleImplementation(py::class_<Type<T>, Args...>& c, std::false_type) {
+    c.def(py::init<Type<U>>(), "Construct from different underlying type");
+}
+
+template<class U, template<class> class Type, class T, class ...Args> void convertibleImplementation(py::class_<Type<T>, Args...>&, std::true_type) {}
+
+template<template<class> class Type, class T, class ...Args> void convertible(py::class_<Type<T>, Args...>& c) {
+    convertibleImplementation<UnsignedInt>(c, std::is_same<T, UnsignedInt>{});
+    convertibleImplementation<Int>(c, std::is_same<T, Int>{});
+    convertibleImplementation<Float>(c, std::is_same<T, Float>{});
+    convertibleImplementation<Double>(c, std::is_same<T, Double>{});
+}
+
 template<class T, class Base> void color(py::class_<T, Base>& c) {
     c
         .def_static("zero_init", []() {
diff --git a/src/python/magnum/math.vectorfloat.cpp b/src/python/magnum/math.vectorfloat.cpp
index 53b8c7a..1a7cd1c 100644
--- a/src/python/magnum/math.vectorfloat.cpp
+++ b/src/python/magnum/math.vectorfloat.cpp
@@ -93,6 +93,18 @@ void mathVectorFloat(py::module& root, py::module& m) {
     py::class_<Color4, Vector4> color4_{root, "Color4", "Color in linear RGBA color space"};
     color(color4_);
     color4(color4_);
+
+    /* Register the integer types as well, only after that register type
+       conversions because they need all the types */
+    mathVectorIntegral(root, m);
+
+    convertible(vector2);
+    convertible(vector3);
+    convertible(vector4);
+    convertible(vector2d);
+    convertible(vector3d);
+    convertible(vector4d);
+    /* Colors are float-only at the moment, thus no conversions */
 }
 
 }
diff --git a/src/python/magnum/math.vectorintegral.cpp b/src/python/magnum/math.vectorintegral.cpp
index e55f6da..87f558a 100644
--- a/src/python/magnum/math.vectorintegral.cpp
+++ b/src/python/magnum/math.vectorintegral.cpp
@@ -78,6 +78,15 @@ void mathVectorIntegral(py::module& root, py::module& m) {
     py::class_<Vector4ui> vector4ui{root, "Vector4ui", "Four-component unsigned integral vector"};
     vectorsIntegral<Int>(m, vector2i, vector3i, vector4i);
     vectorsIntegral<UnsignedInt>(m, vector2ui, vector3ui, vector4ui);
+
+    /* At this point we should have both float and integral types registered,
+       so register type conversions */
+    convertible(vector2i);
+    convertible(vector3i);
+    convertible(vector4i);
+    convertible(vector2ui);
+    convertible(vector3ui);
+    convertible(vector4ui);
 }
 
 }
diff --git a/src/python/magnum/test/test_math.py b/src/python/magnum/test/test_math.py
index 552d8de..48218ac 100644
--- a/src/python/magnum/test/test_math.py
+++ b/src/python/magnum/test/test_math.py
@@ -129,6 +129,10 @@ class Vector(unittest.TestCase):
         self.assertEqual(d, Vector3(1.0, 0.3, 1.1))
         self.assertEqual(e, Vector4d(1.0, 0.3, 1.1, 0.5))
 
+    def test_convert(self):
+        a = Vector2i(Vector2(4.3, 3.1))
+        self.assertEqual(a, Vector2i(4, 3))
+
     def test_static_methods(self):
         self.assertEqual(Vector2.y_scale(5), Vector2(1, 5))
         self.assertEqual(Vector3d.z_axis(-3), Vector3d(0, 0, -3))
@@ -374,6 +378,12 @@ class Matrix(unittest.TestCase):
         self.assertEqual(f[1], Vector2(3.0, 4.0))
         self.assertEqual(f[2], Vector2(5.0, 6.0))
 
+    def test_convert(self):
+        a = Matrix2x3d(Matrix2x3((1.0, 2.0, 3.0),
+                                 (4.0, 5.0, 6.0)))
+        self.assertEqual(a, Matrix2x3d((1.0, 2.0, 3.0),
+                                       (4.0, 5.0, 6.0)))
+
     def test_length(self):
         self.assertEqual(Matrix3x4.__len__(), 3)
         #self.assertEqual(len(Matrix4x3), 4) TODO: Y not?
@@ -459,6 +469,14 @@ class Matrix3_(unittest.TestCase):
         self.assertEqual(d[1], Vector3(4.0, 5.0, 6.0))
         self.assertEqual(d[2], Vector3(7.0, 8.0, 9.0))
 
+    def test_convert(self):
+        a = Matrix3(Matrix3d((1.0, 2.0, 3.0),
+                             (4.0, 5.0, 6.0),
+                             (7.0, 8.0, 9.0)))
+        self.assertEqual(a, Matrix3((1.0, 2.0, 3.0),
+                                    (4.0, 5.0, 6.0),
+                                    (7.0, 8.0, 9.0)))
+
     def test_static_methods(self):
         a = Matrix3.translation((0.0, -1.0))
         self.assertEqual(a[2].xy, Vector2(0.0, -1.0))
@@ -518,6 +536,16 @@ class Matrix4_(unittest.TestCase):
         self.assertEqual(d[2], Vector4(9.0, 10.0, 11.0, 12.0))
         self.assertEqual(d[3], Vector4(13.0, 14.0, 15.0, 16.0))
 
+    def test_convert(self):
+        a = Matrix4d(Matrix4((1.0, 2.0, 3.0, 4.0),
+                             (5.0, 6.0, 7.0, 8.0),
+                             (9.0, 10.0, 11.0, 12.0),
+                             (13.0, 14.0, 15.0, 16.0)))
+        self.assertEqual(a, Matrix4d((1.0, 2.0, 3.0, 4.0),
+                                     (5.0, 6.0, 7.0, 8.0),
+                                     (9.0, 10.0, 11.0, 12.0),
+                                     (13.0, 14.0, 15.0, 16.0)))
+
     def test_static_methods(self):
         a = Matrix4.translation((0.0, -1.0, 2.0))
         self.assertEqual(a[3].xyz, Vector3(0.0, -1.0, 2.0))
@@ -567,6 +595,10 @@ class Quaternion_(unittest.TestCase):
         self.assertEqual(e.vector, Vector3(1.0, 2.0, 3.0))
         self.assertEqual(e.scalar, 4.0)
 
+    def test_convert(self):
+        a = Quaterniond(Quaternion((1.0, 2.0, 3.0), 4.0))
+        self.assertEqual(a, Quaterniond((1.0, 2.0, 3.0), 4.0))
+
     def test_static_methods(self):
         a = Quaternion.rotation(Deg(45.0), Vector3.x_axis())
         self.assertEqual(a, Quaternion((0.382683, 0.0, 0.0), 0.92388))