diff --git a/ext/wayland/gstwaylandsink.c b/ext/wayland/gstwaylandsink.c
index 0b9b89937c..363b31bc48 100644
--- a/ext/wayland/gstwaylandsink.c
+++ b/ext/wayland/gstwaylandsink.c
@@ -333,11 +333,7 @@ gst_wayland_sink_change_state (GstElement * element, GstStateChange transition)
           g_clear_object (&sink->window);
         } else {
           /* remove buffer from surface, show nothing */
-          wl_surface_attach (sink->window->surface, NULL, 0, 0);
-          wl_surface_damage (sink->window->surface, 0, 0,
-              sink->window->surface_width, sink->window->surface_height);
-          wl_surface_commit (sink->window->surface);
-          wl_display_flush (sink->display->display);
+          gst_wl_window_render (sink->window, NULL, NULL);
         }
       }
       break;
@@ -558,6 +554,7 @@ static void
 render_last_buffer (GstWaylandSink * sink)
 {
   GstWlBuffer *wlbuffer;
+  const GstVideoInfo *info = NULL;
   struct wl_surface *surface;
   struct wl_callback *callback;
 
@@ -568,12 +565,11 @@ render_last_buffer (GstWaylandSink * sink)
   callback = wl_surface_frame (surface);
   wl_callback_add_listener (callback, &frame_callback_listener, sink);
 
-  gst_wl_buffer_attach (wlbuffer, sink->window);
-  wl_surface_damage (surface, 0, 0, sink->window->surface_width,
-      sink->window->surface_height);
-
-  wl_surface_commit (surface);
-  wl_display_flush (sink->display->display);
+  if (G_UNLIKELY (sink->video_info_changed)) {
+    info = &sink->video_info;
+    sink->video_info_changed = FALSE;
+  }
+  gst_wl_window_render (sink->window, wlbuffer, info);
 }
 
 static GstFlowReturn
@@ -595,33 +591,19 @@ gst_wayland_sink_render (GstBaseSink * bsink, GstBuffer * buffer)
     gst_video_overlay_prepare_window_handle (GST_VIDEO_OVERLAY (sink));
     g_mutex_lock (&sink->render_lock);
 
-    if (sink->window) {
-      /* inform the window about our caps */
-      gst_wl_window_set_video_info (sink->window, &sink->video_info);
-    } else {
+    if (!sink->window) {
       /* if we were not provided a window, create one ourselves */
       sink->window =
           gst_wl_window_new_toplevel (sink->display, &sink->video_info);
     }
-    sink->video_info_changed = FALSE;
   }
 
   /* drop buffers until we get a frame callback */
   if (g_atomic_int_get (&sink->redraw_pending) == TRUE)
     goto done;
 
-  if (G_UNLIKELY (sink->video_info_changed)) {
-    gst_wl_window_set_video_info (sink->window, &sink->video_info);
-    sink->video_info_changed = FALSE;
-  }
-
-  /* now that we have for sure set the video info on the window, it must have
-   * a valid size, otherwise this means that the application has called
-   * set_window_handle() without calling set_render_rectangle(), which is
-   * absolutely necessary for us.
-   */
-  if (G_UNLIKELY (sink->window->surface_width == 0 ||
-          sink->window->surface_height == 0))
+  /* make sure that the application has called set_render_rectangle() */
+  if (G_UNLIKELY (sink->window->render_rectangle.w == 0))
     goto no_window_size;
 
   wlbuffer = gst_buffer_get_wl_buffer (buffer);
@@ -822,14 +804,14 @@ gst_wayland_sink_begin_geometry_change (GstWaylandVideo * video)
   g_return_if_fail (sink != NULL);
 
   g_mutex_lock (&sink->render_lock);
-  if (!sink->window || !sink->window->subsurface) {
+  if (!sink->window || !sink->window->area_subsurface) {
     g_mutex_unlock (&sink->render_lock);
     GST_INFO_OBJECT (sink,
         "begin_geometry_change called without window, ignoring");
     return;
   }
 
-  wl_subsurface_set_sync (sink->window->subsurface);
+  wl_subsurface_set_sync (sink->window->area_subsurface);
   g_mutex_unlock (&sink->render_lock);
 }
 
@@ -840,14 +822,14 @@ gst_wayland_sink_end_geometry_change (GstWaylandVideo * video)
   g_return_if_fail (sink != NULL);
 
   g_mutex_lock (&sink->render_lock);
-  if (!sink->window || !sink->window->subsurface) {
+  if (!sink->window || !sink->window->area_subsurface) {
     g_mutex_unlock (&sink->render_lock);
     GST_INFO_OBJECT (sink,
         "end_geometry_change called without window, ignoring");
     return;
   }
 
-  wl_subsurface_set_desync (sink->window->subsurface);
+  wl_subsurface_set_desync (sink->window->area_subsurface);
   g_mutex_unlock (&sink->render_lock);
 }
 
diff --git a/ext/wayland/wlbuffer.c b/ext/wayland/wlbuffer.c
index c2bf3208b2..d93fedb7ff 100644
--- a/ext/wayland/wlbuffer.c
+++ b/ext/wayland/wlbuffer.c
@@ -120,7 +120,7 @@ static const struct wl_buffer_listener buffer_listener = {
   buffer_release
 };
 
-void
+GstWlBuffer *
 gst_buffer_add_wl_buffer (GstBuffer * gstbuffer, struct wl_buffer *wlbuffer,
     GstWlDisplay * display)
 {
@@ -137,6 +137,8 @@ gst_buffer_add_wl_buffer (GstBuffer * gstbuffer, struct wl_buffer *wlbuffer,
 
   gst_mini_object_set_qdata ((GstMiniObject *) gstbuffer,
       gst_wl_buffer_qdata_quark (), self, g_object_unref);
+
+  return self;
 }
 
 GstWlBuffer *
@@ -172,11 +174,11 @@ gst_wl_buffer_force_release_and_unref (GstWlBuffer * self)
 }
 
 void
-gst_wl_buffer_attach (GstWlBuffer * self, GstWlWindow * target)
+gst_wl_buffer_attach (GstWlBuffer * self, struct wl_surface *surface)
 {
   g_return_if_fail (self->used_by_compositor == FALSE);
 
-  wl_surface_attach (target->surface, self->wlbuffer, 0, 0);
+  wl_surface_attach (surface, self->wlbuffer, 0, 0);
 
   /* Add a reference to the buffer. This represents the fact that
    * the compositor is using the buffer and it should not return
diff --git a/ext/wayland/wlbuffer.h b/ext/wayland/wlbuffer.h
index b434df469c..cbb50f719a 100644
--- a/ext/wayland/wlbuffer.h
+++ b/ext/wayland/wlbuffer.h
@@ -21,7 +21,7 @@
 #ifndef __GST_WL_BUFFER_H__
 #define __GST_WL_BUFFER_H__
 
-#include "wlwindow.h"
+#include "wldisplay.h"
 
 G_BEGIN_DECLS
 
@@ -54,13 +54,13 @@ struct _GstWlBufferClass
 
 GType gst_wl_buffer_get_type (void);
 
-void gst_buffer_add_wl_buffer (GstBuffer * gstbuffer,
+GstWlBuffer * gst_buffer_add_wl_buffer (GstBuffer * gstbuffer,
     struct wl_buffer * wlbuffer, GstWlDisplay * display);
 GstWlBuffer * gst_buffer_get_wl_buffer (GstBuffer * gstbuffer);
 
 void gst_wl_buffer_force_release_and_unref (GstWlBuffer * self);
 
-void gst_wl_buffer_attach (GstWlBuffer * self, GstWlWindow * target);
+void gst_wl_buffer_attach (GstWlBuffer * self, struct wl_surface *surface);
 
 G_END_DECLS
 
diff --git a/ext/wayland/wlvideoformat.h b/ext/wayland/wlvideoformat.h
index 833b98db34..e8ec7ae129 100644
--- a/ext/wayland/wlvideoformat.h
+++ b/ext/wayland/wlvideoformat.h
@@ -24,7 +24,7 @@
 #ifndef __GST_WL_VIDEO_FORMAT_H__
 #define __GST_WL_VIDEO_FORMAT_H__
 
-#include <wayland-client.h>
+#include <wayland-client-protocol.h>
 #include <gst/video/video.h>
 
 G_BEGIN_DECLS
diff --git a/ext/wayland/wlwindow.c b/ext/wayland/wlwindow.c
index f58df09654..a4da2c6624 100644
--- a/ext/wayland/wlwindow.c
+++ b/ext/wayland/wlwindow.c
@@ -25,6 +25,8 @@
 #endif
 
 #include "wlwindow.h"
+#include "wlshmallocator.h"
+#include "wlbuffer.h"
 
 GST_DEBUG_CATEGORY_EXTERN (gstwayland_debug);
 #define GST_CAT_DEFAULT gstwayland_debug
@@ -78,12 +80,15 @@ gst_wl_window_finalize (GObject * gobject)
     wl_shell_surface_destroy (self->shell_surface);
   }
 
-  if (self->subsurface) {
-    wl_subsurface_destroy (self->subsurface);
-  }
+  wl_viewport_destroy (self->video_viewport);
+  wl_subsurface_destroy (self->video_subsurface);
+  wl_surface_destroy (self->video_surface);
 
-  wl_viewport_destroy (self->viewport);
-  wl_surface_destroy (self->surface);
+  if (self->area_subsurface) {
+    wl_subsurface_destroy (self->area_subsurface);
+  }
+  wl_viewport_destroy (self->area_viewport);
+  wl_surface_destroy (self->area_surface);
 
   g_clear_object (&self->display);
 
@@ -91,44 +96,84 @@ gst_wl_window_finalize (GObject * gobject)
 }
 
 static GstWlWindow *
-gst_wl_window_new_internal (GstWlDisplay * display, struct wl_surface *surface)
+gst_wl_window_new_internal (GstWlDisplay * display)
 {
   GstWlWindow *window;
+  GstVideoInfo info;
+  GstBuffer *buf;
+  GstMapInfo mapinfo;
+  struct wl_buffer *wlbuf;
+  GstWlBuffer *gwlbuf;
   struct wl_region *region;
 
-  g_return_val_if_fail (surface != NULL, NULL);
-
   window = g_object_new (GST_TYPE_WL_WINDOW, NULL);
   window->display = g_object_ref (display);
-  window->surface = surface;
 
-  /* make sure the surface runs on our local queue */
-  wl_proxy_set_queue ((struct wl_proxy *) surface, display->queue);
+  window->area_surface = wl_compositor_create_surface (display->compositor);
+  window->video_surface = wl_compositor_create_surface (display->compositor);
 
-  window->viewport = wl_scaler_get_viewport (display->scaler, window->surface);
+  wl_proxy_set_queue ((struct wl_proxy *) window->area_surface, display->queue);
+  wl_proxy_set_queue ((struct wl_proxy *) window->video_surface,
+      display->queue);
+
+  /* embed video_surface in area_surface */
+  window->video_subsurface =
+      wl_subcompositor_get_subsurface (display->subcompositor,
+      window->video_surface, window->area_surface);
+  wl_subsurface_set_desync (window->video_subsurface);
+
+  window->area_viewport = wl_scaler_get_viewport (display->scaler,
+      window->area_surface);
+  window->video_viewport = wl_scaler_get_viewport (display->scaler,
+      window->video_surface);
+
+  /* draw the area_subsurface */
+  gst_video_info_set_format (&info,
+      /* we want WL_SHM_FORMAT_XRGB8888 */
+#if G_BYTE_ORDER == G_BIG_ENDIAN
+      GST_VIDEO_FORMAT_xRGB,
+#else
+      GST_VIDEO_FORMAT_BGRx,
+#endif
+      1, 1);
+
+  buf = gst_buffer_new_allocate (gst_wl_shm_allocator_get (), info.size, NULL);
+  gst_buffer_map (buf, &mapinfo, GST_MAP_WRITE);
+  *((guint32 *) mapinfo.data) = 0;      /* paint it black */
+  gst_buffer_unmap (buf, &mapinfo);
+  wlbuf =
+      gst_wl_shm_memory_construct_wl_buffer (gst_buffer_peek_memory (buf, 0),
+      display, &info);
+  gwlbuf = gst_buffer_add_wl_buffer (buf, wlbuf, display);
+  gst_wl_buffer_attach (gwlbuf, window->area_surface);
+
+  /* at this point, the GstWlBuffer keeps the buffer
+   * alive and will free it on wl_buffer::release */
+  gst_buffer_unref (buf);
 
   /* do not accept input */
   region = wl_compositor_create_region (display->compositor);
-  wl_surface_set_input_region (surface, region);
+  wl_surface_set_input_region (window->area_surface, region);
+  wl_region_destroy (region);
+
+  region = wl_compositor_create_region (display->compositor);
+  wl_surface_set_input_region (window->video_surface, region);
   wl_region_destroy (region);
 
   return window;
 }
 
 GstWlWindow *
-gst_wl_window_new_toplevel (GstWlDisplay * display, GstVideoInfo * video_info)
+gst_wl_window_new_toplevel (GstWlDisplay * display, const GstVideoInfo * info)
 {
   GstWlWindow *window;
+  gint width;
 
-  window = gst_wl_window_new_internal (display,
-      wl_compositor_create_surface (display->compositor));
-
-  gst_wl_window_set_video_info (window, video_info);
-  gst_wl_window_set_render_rectangle (window, 0, 0, window->video_width,
-      window->video_height);
+  window = gst_wl_window_new_internal (display);
 
+  /* go toplevel */
   window->shell_surface = wl_shell_get_shell_surface (display->shell,
-      window->surface);
+      window->area_surface);
 
   if (window->shell_surface) {
     wl_shell_surface_add_listener (window->shell_surface,
@@ -141,6 +186,11 @@ gst_wl_window_new_toplevel (GstWlDisplay * display, GstVideoInfo * video_info)
     return NULL;
   }
 
+  /* set the initial size to be the same as the reported video size */
+  width =
+      gst_util_uint64_scale_int_round (info->width, info->par_n, info->par_d);
+  gst_wl_window_set_render_rectangle (window, 0, 0, width, info->height);
+
   return window;
 }
 
@@ -149,13 +199,13 @@ gst_wl_window_new_in_surface (GstWlDisplay * display,
     struct wl_surface * parent)
 {
   GstWlWindow *window;
+  window = gst_wl_window_new_internal (display);
 
-  window = gst_wl_window_new_internal (display,
-      wl_compositor_create_surface (display->compositor));
-
-  window->subsurface = wl_subcompositor_get_subsurface (display->subcompositor,
-      window->surface, parent);
-  wl_subsurface_set_desync (window->subsurface);
+  /* embed in parent */
+  window->area_subsurface =
+      wl_subcompositor_get_subsurface (display->subcompositor,
+      window->area_surface, parent);
+  wl_subsurface_set_desync (window->area_subsurface);
 
   return window;
 }
@@ -173,7 +223,7 @@ gst_wl_window_get_wl_surface (GstWlWindow * window)
 {
   g_return_val_if_fail (window != NULL, NULL);
 
-  return window->surface;
+  return window->video_surface;
 }
 
 gboolean
@@ -185,22 +235,21 @@ gst_wl_window_is_toplevel (GstWlWindow * window)
 }
 
 static void
-gst_wl_window_resize_internal (GstWlWindow * window, gboolean commit)
+gst_wl_window_resize_video_surface (GstWlWindow * window, gboolean commit)
 {
   GstVideoRectangle src, res;
 
+  /* center the video_subsurface inside area_subsurface */
   src.w = window->video_width;
   src.h = window->video_height;
   gst_video_sink_center_rect (src, window->render_rectangle, &res, TRUE);
 
-  if (window->subsurface)
-    wl_subsurface_set_position (window->subsurface,
-        window->render_rectangle.x + res.x, window->render_rectangle.y + res.y);
-  wl_viewport_set_destination (window->viewport, res.w, res.h);
+  wl_subsurface_set_position (window->video_subsurface, res.x, res.y);
+  wl_viewport_set_destination (window->video_viewport, res.w, res.h);
 
   if (commit) {
-    wl_surface_damage (window->surface, 0, 0, res.w, res.h);
-    wl_surface_commit (window->surface);
+    wl_surface_damage (window->video_surface, 0, 0, res.w, res.h);
+    wl_surface_commit (window->video_surface);
   }
 
   /* this is saved for use in wl_surface_damage */
@@ -209,16 +258,37 @@ gst_wl_window_resize_internal (GstWlWindow * window, gboolean commit)
 }
 
 void
-gst_wl_window_set_video_info (GstWlWindow * window, GstVideoInfo * info)
+gst_wl_window_render (GstWlWindow * window, GstWlBuffer * buffer,
+    const GstVideoInfo * info)
 {
-  g_return_if_fail (window != NULL);
+  if (G_UNLIKELY (info)) {
+    window->video_width =
+        gst_util_uint64_scale_int_round (info->width, info->par_n, info->par_d);
+    window->video_height = info->height;
 
-  window->video_width =
-      gst_util_uint64_scale_int_round (info->width, info->par_n, info->par_d);
-  window->video_height = info->height;
+    wl_subsurface_set_sync (window->video_subsurface);
+    gst_wl_window_resize_video_surface (window, FALSE);
+  }
 
-  if (window->render_rectangle.w != 0)
-    gst_wl_window_resize_internal (window, FALSE);
+  if (G_LIKELY (buffer))
+    gst_wl_buffer_attach (buffer, window->video_surface);
+  else
+    wl_surface_attach (window->video_surface, NULL, 0, 0);
+
+  wl_surface_damage (window->video_surface, 0, 0, window->surface_width,
+      window->surface_height);
+  wl_surface_commit (window->video_surface);
+
+  if (G_UNLIKELY (info)) {
+    /* commit also the parent (area_surface) in order to change
+     * the position of the video_subsurface */
+    wl_surface_damage (window->area_surface, 0, 0, window->render_rectangle.w,
+        window->render_rectangle.h);
+    wl_surface_commit (window->area_surface);
+    wl_subsurface_set_desync (window->video_subsurface);
+  }
+
+  wl_display_flush (window->display->display);
 }
 
 void
@@ -232,6 +302,21 @@ gst_wl_window_set_render_rectangle (GstWlWindow * window, gint x, gint y,
   window->render_rectangle.w = w;
   window->render_rectangle.h = h;
 
+  /* position the area inside the parent - needs a parent commit to apply */
+  if (window->area_subsurface)
+    wl_subsurface_set_position (window->area_subsurface, x, y);
+
+  /* change the size of the area */
+  wl_viewport_set_destination (window->area_viewport, w, h);
+
+  if (window->video_width != 0) {
+    wl_subsurface_set_sync (window->video_subsurface);
+    gst_wl_window_resize_video_surface (window, TRUE);
+  }
+
+  wl_surface_damage (window->area_surface, 0, 0, w, h);
+  wl_surface_commit (window->area_surface);
+
   if (window->video_width != 0)
-    gst_wl_window_resize_internal (window, TRUE);
+    wl_subsurface_set_desync (window->video_subsurface);
 }
diff --git a/ext/wayland/wlwindow.h b/ext/wayland/wlwindow.h
index f4cb36083f..e22cb261b5 100644
--- a/ext/wayland/wlwindow.h
+++ b/ext/wayland/wlwindow.h
@@ -22,6 +22,7 @@
 #define __GST_WL_WINDOW_H__
 
 #include "wldisplay.h"
+#include "wlbuffer.h"
 #include <gst/video/video.h>
 
 G_BEGIN_DECLS
@@ -41,16 +42,19 @@ struct _GstWlWindow
   GObject parent_instance;
 
   GstWlDisplay *display;
-  struct wl_surface *surface;
-  struct wl_subsurface *subsurface;
-  struct wl_viewport *viewport;
+  struct wl_surface *area_surface;
+  struct wl_subsurface *area_subsurface;
+  struct wl_viewport *area_viewport;
+  struct wl_surface *video_surface;
+  struct wl_subsurface *video_subsurface;
+  struct wl_viewport *video_viewport;
   struct wl_shell_surface *shell_surface;
 
-  /* the size of the destination area where we are overlaying our subsurface */
+  /* the size and position of the area_(sub)surface */
   GstVideoRectangle render_rectangle;
   /* the size of the video in the buffers */
   gint video_width, video_height;
-  /* the size of the (sub)surface */
+  /* the size of the video_(sub)surface */
   gint surface_width, surface_height;
 };
 
@@ -62,7 +66,7 @@ struct _GstWlWindowClass
 GType gst_wl_window_get_type (void);
 
 GstWlWindow *gst_wl_window_new_toplevel (GstWlDisplay * display,
-        GstVideoInfo * video_info);
+        const GstVideoInfo * info);
 GstWlWindow *gst_wl_window_new_in_surface (GstWlDisplay * display,
         struct wl_surface * parent);
 
@@ -70,8 +74,8 @@ GstWlDisplay *gst_wl_window_get_display (GstWlWindow * window);
 struct wl_surface *gst_wl_window_get_wl_surface (GstWlWindow * window);
 gboolean gst_wl_window_is_toplevel (GstWlWindow *window);
 
-/* functions to manipulate the size on non-toplevel windows */
-void gst_wl_window_set_video_info (GstWlWindow * window, GstVideoInfo * info);
+void gst_wl_window_render (GstWlWindow * window, GstWlBuffer * buffer,
+        const GstVideoInfo * info);
 void gst_wl_window_set_render_rectangle (GstWlWindow * window, gint x, gint y,
         gint w, gint h);