diff --git a/sys/vdpau/gstvdputils.c b/sys/vdpau/gstvdputils.c
index 8ac4f4217a..fc8d46e641 100644
--- a/sys/vdpau/gstvdputils.c
+++ b/sys/vdpau/gstvdputils.c
@@ -20,8 +20,8 @@
 
 #include "gstvdputils.h"
 
-GstCaps *
-gst_vdp_get_video_caps (GstVdpDevice * device, gint chroma_format)
+static GstCaps *
+gst_vdp_get_allowed_yuv_caps (GstVdpDevice * device)
 {
   GstCaps *caps;
   gint i;
@@ -32,9 +32,6 @@ gst_vdp_get_video_caps (GstVdpDevice * device, gint chroma_format)
     VdpBool is_supported;
     guint32 max_w, max_h;
 
-    if (chroma_format != chroma_types[i])
-      continue;
-
     status =
         device->vdp_video_surface_query_capabilities (device->device,
         chroma_types[i], &is_supported, &max_w, &max_h);
@@ -80,10 +77,154 @@ gst_vdp_get_video_caps (GstVdpDevice * device, gint chroma_format)
   }
 
 error:
-  if (gst_caps_is_empty (caps)) {
-    gst_caps_unref (caps);
-    return NULL;
-  }
-
   return caps;
 }
+
+GstCaps *
+gst_vdp_video_to_yuv_caps (GstCaps * caps)
+{
+  GstCaps *new_caps, *allowed_caps, *result;
+  gint i;
+  GstStructure *structure;
+  const GValue *value;
+  GstVdpDevice *device = NULL;
+
+  new_caps = gst_caps_new_empty ();
+
+  for (i = 0; i < gst_caps_get_size (caps); i++) {
+    gint chroma_type;
+    GSList *fourcc = NULL, *iter;
+
+    structure = gst_caps_get_structure (caps, i);
+
+    if (gst_structure_get_int (structure, "chroma-type", &chroma_type)) {
+      /* calculate fourcc from chroma_type */
+      for (i = 0; i < N_FORMATS; i++) {
+        if (formats[i].chroma_type == chroma_type) {
+          fourcc = g_slist_append (fourcc, GINT_TO_POINTER (formats[i].fourcc));
+        }
+      }
+    } else {
+      for (i = 0; i < N_FORMATS; i++) {
+        fourcc = g_slist_append (fourcc, GINT_TO_POINTER (formats[i].fourcc));
+      }
+    }
+
+    for (iter = fourcc; iter; iter = iter->next) {
+      GstStructure *new_struct = gst_structure_copy (structure);
+
+      gst_structure_set_name (new_struct, "video/x-raw-yuv");
+      gst_structure_remove_field (new_struct, "chroma-type");
+      gst_structure_remove_field (new_struct, "device");
+      gst_structure_set (new_struct, "format", GST_TYPE_FOURCC,
+          GPOINTER_TO_INT (iter->data), NULL);
+
+      gst_caps_append_structure (new_caps, new_struct);
+    }
+
+    g_slist_free (fourcc);
+  }
+  structure = gst_caps_get_structure (caps, 0);
+
+  value = gst_structure_get_value (structure, "device");
+  if (value)
+    device = g_value_get_object (value);
+
+  if (device) {
+    allowed_caps = gst_vdp_get_allowed_yuv_caps (device);
+    result = gst_caps_intersect (new_caps, allowed_caps);
+
+    gst_caps_unref (new_caps);
+    gst_caps_unref (allowed_caps);
+  } else
+    result = new_caps;
+
+  return result;
+}
+
+static GstCaps *
+gst_vdp_get_allowed_video_caps (GstVdpDevice * device)
+{
+  GstCaps *caps;
+  gint i;
+
+  caps = gst_caps_new_empty ();
+  for (i = 0; i < N_CHROMA_TYPES; i++) {
+    VdpStatus status;
+    VdpBool is_supported;
+    guint32 max_w, max_h;
+
+    status =
+        device->vdp_video_surface_query_capabilities (device->device,
+        chroma_types[i], &is_supported, &max_w, &max_h);
+
+    if (status != VDP_STATUS_OK && status != VDP_STATUS_INVALID_CHROMA_TYPE) {
+      GST_ERROR_OBJECT (device,
+          "Could not get query VDPAU video surface capabilites, "
+          "Error returned from vdpau was: %s",
+          device->vdp_get_error_string (status));
+
+      goto error;
+    }
+
+    if (is_supported) {
+      GstCaps *format_caps;
+
+      format_caps = gst_caps_new_simple ("video/x-vdpau-video",
+          "chroma-type", G_TYPE_INT, chroma_types[i],
+          "width", GST_TYPE_INT_RANGE, 1, max_w,
+          "height", GST_TYPE_INT_RANGE, 1, max_h, NULL);
+      gst_caps_append (caps, format_caps);
+    }
+  }
+
+error:
+  return caps;
+}
+
+GstCaps *
+gst_vdp_yuv_to_video_caps (GstCaps * caps, GstVdpDevice * device)
+{
+  GstCaps *new_caps, *result;
+  gint i;
+
+  new_caps = gst_caps_copy (caps);
+  for (i = 0; i < gst_caps_get_size (new_caps); i++) {
+    GstStructure *structure = gst_caps_get_structure (new_caps, i);
+    guint32 fourcc;
+
+    if (gst_structure_get_fourcc (structure, "format", &fourcc)) {
+      gint chroma_type = -1;
+
+      /* calculate chroma type from fourcc */
+      for (i = 0; i < N_FORMATS; i++) {
+        if (formats[i].fourcc == fourcc) {
+          chroma_type = formats[i].chroma_type;
+          break;
+        }
+      }
+      gst_structure_remove_field (structure, "format");
+      gst_structure_set (structure, "chroma-type", G_TYPE_INT, chroma_type,
+          NULL);
+    } else
+      gst_structure_set (structure, "chroma-type", GST_TYPE_INT_RANGE, 0, 2,
+          NULL);
+
+    gst_structure_set_name (structure, "video/x-vdpau-video");
+    if (device)
+      gst_structure_set (structure, "device", G_TYPE_OBJECT, device, NULL);
+  }
+
+  if (device) {
+    GstCaps *allowed_caps;
+
+    allowed_caps = gst_vdp_get_allowed_video_caps (device);
+    result = gst_caps_intersect (new_caps, allowed_caps);
+
+    gst_caps_unref (new_caps);
+    gst_caps_unref (allowed_caps);
+  } else
+    result = new_caps;
+
+  return result;
+}
diff --git a/sys/vdpau/gstvdputils.h b/sys/vdpau/gstvdputils.h
index cb6a965f14..530a4adea8 100644
--- a/sys/vdpau/gstvdputils.h
+++ b/sys/vdpau/gstvdputils.h
@@ -76,6 +76,7 @@ static const VdpauFormats formats[N_FORMATS] = {
       }
 };
 
-GstCaps *gst_vdp_get_video_caps (GstVdpDevice *device, gint chroma_format);
+GstCaps *gst_vdp_video_to_yuv_caps (GstCaps *caps);
+GstCaps *gst_vdp_yuv_to_video_caps (GstCaps *caps, GstVdpDevice *device);
 
 #endif /* _GST_VDP_UTILS_H_ */
\ No newline at end of file
diff --git a/sys/vdpau/gstvdpvideoyuv.c b/sys/vdpau/gstvdpvideoyuv.c
index 7b3aa9ea87..1810f9e06d 100644
--- a/sys/vdpau/gstvdpvideoyuv.c
+++ b/sys/vdpau/gstvdpvideoyuv.c
@@ -51,12 +51,11 @@ GST_STATIC_PAD_TEMPLATE (GST_BASE_TRANSFORM_SINK_NAME,
     GST_STATIC_CAPS (GST_VDP_VIDEO_CAPS));
 
 static GstStaticPadTemplate src_template =
-GST_STATIC_PAD_TEMPLATE (GST_BASE_TRANSFORM_SRC_NAME,
+    GST_STATIC_PAD_TEMPLATE (GST_BASE_TRANSFORM_SRC_NAME,
     GST_PAD_SRC,
     GST_PAD_ALWAYS,
-    GST_STATIC_CAPS ("video/x-raw-yuv, "
-        "framerate = (fraction) [ 0, MAX ], "
-        "width = (int) [ 1, MAX ], " "height = (int) [ 1, MAX ]"));
+    GST_STATIC_CAPS (GST_VIDEO_CAPS_YUV ("I420") ";"
+        GST_VIDEO_CAPS_YUV ("YV12") ";" GST_VIDEO_CAPS_YUV ("NV12")));
 
 #define DEBUG_INIT(bla) \
     GST_DEBUG_CATEGORY_INIT (gst_vdp_video_yuv_debug, "vdpauvideoyuv", 0, "VDPAU VdpSurface to YUV");
@@ -254,67 +253,15 @@ GstCaps *
 gst_vdp_video_yuv_transform_caps (GstBaseTransform * trans,
     GstPadDirection direction, GstCaps * caps)
 {
-  GstCaps *result = NULL;
+  GstCaps *result;
 
-  if (direction == GST_PAD_SINK) {
-    GstCaps *new_caps, *allowed_caps = NULL;
-    gint i;
-    GstStructure *structure;
-    gint chroma_type;
-    const GValue *value;
-    GstVdpDevice *device = NULL;
+  if (direction == GST_PAD_SINK)
+    result = gst_vdp_video_to_yuv_caps (caps);
 
-    new_caps = gst_caps_new_empty ();
+  else if (direction == GST_PAD_SRC)
+    result = gst_vdp_yuv_to_video_caps (caps, NULL);
 
-    for (i = 0; i < gst_caps_get_size (caps); i++) {
-      GSList *fourcc = NULL, *iter;
-
-      structure = gst_caps_get_structure (caps, i);
-      gst_structure_get_int (structure, "chroma-type", &chroma_type);
-      /* calculate fourcc from chroma_type */
-      for (i = 0; i < N_FORMATS; i++) {
-        if (formats[i].chroma_type == chroma_type) {
-          fourcc = g_slist_append (fourcc, GINT_TO_POINTER (formats[i].fourcc));
-        }
-      }
-
-      for (iter = fourcc; iter; iter = iter->next) {
-        GstStructure *new_struct = gst_structure_copy (structure);
-
-        gst_structure_set_name (new_struct, "video/x-raw-yuv");
-        gst_structure_remove_field (new_struct, "chroma-type");
-        gst_structure_remove_field (new_struct, "device");
-        gst_structure_set (new_struct, "format", GST_TYPE_FOURCC,
-            GPOINTER_TO_INT (iter->data), NULL);
-
-        gst_caps_append_structure (new_caps, new_struct);
-      }
-
-      g_slist_free (fourcc);
-    }
-    structure = gst_caps_get_structure (caps, 0);
-
-    gst_structure_get_int (structure, "chroma-type", &chroma_type);
-    value = gst_structure_get_value (structure, "device");
-    if (value)
-      device = g_value_get_object (value);
-
-    if (device)
-      allowed_caps = gst_vdp_get_video_caps (device, chroma_type);
-    else
-      allowed_caps = gst_static_pad_template_get_caps (&src_template);
-
-    result = gst_caps_intersect (new_caps, allowed_caps);
-    gst_caps_unref (new_caps);
-    gst_caps_unref (allowed_caps);
-
-    GST_LOG ("transformed %" GST_PTR_FORMAT " to %" GST_PTR_FORMAT, caps,
-        result);
-
-  } else if (direction == GST_PAD_SRC) {
-    /* FIXME: upstream negotiation */
-    result = gst_static_pad_template_get_caps (&sink_template);
-  }
+  GST_LOG ("transformed %" GST_PTR_FORMAT " to %" GST_PTR_FORMAT, caps, result);
 
   return result;
 }
diff --git a/sys/vdpau/gstvdpyuvvideo.c b/sys/vdpau/gstvdpyuvvideo.c
index f602965d3e..04c5fe5a0f 100644
--- a/sys/vdpau/gstvdpyuvvideo.c
+++ b/sys/vdpau/gstvdpyuvvideo.c
@@ -235,44 +235,16 @@ gst_vdp_yuv_video_transform_caps (GstBaseTransform * trans,
     GstPadDirection direction, GstCaps * caps)
 {
   GstVdpYUVVideo *yuv_video = GST_VDP_YUV_VIDEO (trans);
-  GstCaps *result = NULL;
+  GstCaps *result;
 
   if (direction == GST_PAD_SINK) {
-    gint i;
-
-    /* Intersect with the allowed caps */
-    if (yuv_video->sink_caps)
-      result = gst_caps_intersect (caps, yuv_video->sink_caps);
-    else
-      result = gst_caps_copy (caps);
-
-    for (i = 0; i < gst_caps_get_size (result); i++) {
-      GstStructure *structure = gst_caps_get_structure (result, i);
-      guint32 fourcc;
-      gint chroma_type;
-
-      gst_structure_get_fourcc (structure, "format", &fourcc);
-      /* calculate chroma type from fourcc */
-      for (i = 0; i < N_FORMATS; i++) {
-        if (formats[i].fourcc == fourcc) {
-          chroma_type = formats[i].chroma_type;
-          break;
-        }
-      }
-
-      gst_structure_set_name (structure, "video/x-vdpau-video");
-      gst_structure_remove_field (structure, "format");
-      gst_structure_set (structure, "chroma-type", G_TYPE_INT, chroma_type,
-          "device", G_TYPE_OBJECT, yuv_video->device, NULL);
-    }
-    gst_caps_do_simplify (result);
-    GST_LOG ("transformed %" GST_PTR_FORMAT " to %" GST_PTR_FORMAT, caps,
-        result);
+    result = gst_vdp_yuv_to_video_caps (caps, yuv_video->device);
   } else if (direction == GST_PAD_SRC) {
-    /* FIXME: upstream negotiation */
-    result = gst_static_pad_template_get_caps (&sink_template);
+    result = gst_vdp_video_to_yuv_caps (caps);
   }
 
+  GST_LOG ("transformed %" GST_PTR_FORMAT " to %" GST_PTR_FORMAT, caps, result);
+
   return result;
 }
 
@@ -284,7 +256,6 @@ gst_vdp_yuv_video_start (GstBaseTransform * trans)
   yuv_video->device = gst_vdp_get_device (yuv_video->display);
   if (!yuv_video->device)
     return FALSE;
-  yuv_video->sink_caps = gst_vdp_get_video_caps (yuv_video->device, -1);
 
   return TRUE;
 }
@@ -295,8 +266,6 @@ gst_vdp_yuv_video_stop (GstBaseTransform * trans)
   GstVdpYUVVideo *yuv_video = GST_VDP_YUV_VIDEO (trans);
 
   g_object_unref (yuv_video->device);
-  gst_caps_unref (yuv_video->sink_caps);
-  yuv_video->sink_caps = NULL;
 
   return TRUE;
 }
@@ -390,7 +359,5 @@ gst_vdp_yuv_video_class_init (GstVdpYUVVideoClass * klass)
 static void
 gst_vdp_yuv_video_init (GstVdpYUVVideo * yuv_video, GstVdpYUVVideoClass * klass)
 {
-  yuv_video->sink_caps = NULL;
-
   yuv_video->display = NULL;
 }
diff --git a/sys/vdpau/gstvdpyuvvideo.h b/sys/vdpau/gstvdpyuvvideo.h
index 891756afce..8851ce443a 100644
--- a/sys/vdpau/gstvdpyuvvideo.h
+++ b/sys/vdpau/gstvdpyuvvideo.h
@@ -41,8 +41,6 @@ typedef struct _GstVdpYUVVideoClass GstVdpYUVVideoClass;
 struct _GstVdpYUVVideo {
   GstBaseTransform trans;
 
-  GstCaps *sink_caps;
-
   gchar *display;
   GstVdpDevice *device;