diff --git a/GL/texture.c b/GL/texture.c
index 18c1db8..e1f2d3a 100644
--- a/GL/texture.c
+++ b/GL/texture.c
@@ -170,24 +170,28 @@ GLboolean APIENTRY glIsTexture(GLuint texture) {
     return (named_array_used(&TEXTURE_OBJECTS, texture)) ? GL_TRUE : GL_FALSE;
 }
 
+static void _glInitializeTextureObject(TextureObject* txr, unsigned int id) {
+    txr->index = id;
+    txr->width = txr->height = 0;
+    txr->mipmap = 0;
+    txr->uv_clamp = 0;
+    txr->env = PVR_TXRENV_MODULATEALPHA;
+    txr->data = NULL;
+    txr->mipmapCount = 0;
+    txr->minFilter = GL_NEAREST;
+    txr->magFilter = GL_NEAREST;
+    txr->palette = NULL;
+    txr->isCompressed = GL_FALSE;
+    txr->isPaletted = GL_FALSE;
+}
+
 void APIENTRY glGenTextures(GLsizei n, GLuint *textures) {
     TRACE();
 
     while(n--) {
         GLuint id = 0;
         TextureObject* txr = (TextureObject*) named_array_alloc(&TEXTURE_OBJECTS, &id);
-        txr->index = id;
-        txr->width = txr->height = 0;
-        txr->mipmap = 0;
-        txr->uv_clamp = 0;
-        txr->env = PVR_TXRENV_MODULATEALPHA;
-        txr->data = NULL;
-        txr->mipmapCount = 0;
-        txr->minFilter = GL_NEAREST;
-        txr->magFilter = GL_NEAREST;
-        txr->palette = NULL;
-        txr->isCompressed = GL_FALSE;
-        txr->isPaletted = GL_FALSE;
+        _glInitializeTextureObject(txr, id);
 
         *textures = id;
 
@@ -236,6 +240,13 @@ void APIENTRY glBindTexture(GLenum  target, GLuint texture) {
         return;
     }
 
+    /* If this didn't come from glGenTextures, then we should initialize the
+     * texture the first time it's bound */
+    if(!named_array_used(&TEXTURE_OBJECTS, texture)) {
+        TextureObject* txr = named_array_reserve(&TEXTURE_OBJECTS, texture);
+        _glInitializeTextureObject(txr, texture);
+    }
+
     if(texture) {
         TEXTURE_UNITS[ACTIVE_TEXTURE] = (TextureObject*) named_array_get(&TEXTURE_OBJECTS, texture);
     } else {
diff --git a/containers/named_array.c b/containers/named_array.c
index 7ddfc7e..4ae07bd 100644
--- a/containers/named_array.c
+++ b/containers/named_array.c
@@ -1,6 +1,7 @@
 #include <stdlib.h>
 #include <math.h>
 #include <string.h>
+#include <assert.h>
 
 #ifndef __APPLE__
 #include <malloc.h>
@@ -57,6 +58,23 @@ void* named_array_alloc(NamedArray* array, unsigned int* new_id) {
     return NULL;
 }
 
+void* named_array_reserve(NamedArray* array, unsigned int id) {
+    if(!named_array_used(array, id)) {
+        unsigned int j = id % 8;
+        unsigned int i = (id - j) / 8;
+
+        assert(!named_array_used(array, id));
+        array->used_markers[i] |= (unsigned char) 1 << j;
+        assert(named_array_used(array, id));
+
+        unsigned char* ptr = &array->elements[(id - 1) * array->element_size];
+        memset(ptr, 0, array->element_size);
+        return ptr;
+    }
+
+    return named_array_get(array, id);
+}
+
 void named_array_release(NamedArray* array, unsigned int new_id) {
     new_id--;
 
diff --git a/containers/named_array.h b/containers/named_array.h
index ecf4d7f..a0f6c97 100644
--- a/containers/named_array.h
+++ b/containers/named_array.h
@@ -15,7 +15,10 @@ typedef struct {
 
 void named_array_init(NamedArray* array, unsigned int element_size, unsigned int max_elements);
 char named_array_used(NamedArray* array, unsigned int id);
+
 void* named_array_alloc(NamedArray* array, unsigned int* new_id);
+void* named_array_reserve(NamedArray* array, unsigned int id);
+
 void named_array_release(NamedArray* array, unsigned int new_id);
 void* named_array_get(NamedArray* array, unsigned int id);
 void named_array_cleanup(NamedArray* array);