diff --git a/ext/theora/gsttheoradec.c b/ext/theora/gsttheoradec.c
index d3f7dbd679..52d8f9599e 100644
--- a/ext/theora/gsttheoradec.c
+++ b/ext/theora/gsttheoradec.c
@@ -50,7 +50,6 @@
 #define GST_CAT_DEFAULT theoradec_debug
 GST_DEBUG_CATEGORY_STATIC (GST_CAT_DEFAULT);
 
-#define THEORA_DEF_CROP         TRUE
 #define THEORA_DEF_TELEMETRY_MV 0
 #define THEORA_DEF_TELEMETRY_MBMODE 0
 #define THEORA_DEF_TELEMETRY_QI 0
@@ -59,7 +58,6 @@ GST_DEBUG_CATEGORY_STATIC (GST_CAT_DEFAULT);
 enum
 {
   PROP_0,
-  PROP_CROP,
   PROP_TELEMETRY_MV,
   PROP_TELEMETRY_MBMODE,
   PROP_TELEMETRY_QI,
@@ -128,11 +126,6 @@ gst_theora_dec_class_init (GstTheoraDecClass * klass)
   gobject_class->set_property = theora_dec_set_property;
   gobject_class->get_property = theora_dec_get_property;
 
-  g_object_class_install_property (gobject_class, PROP_CROP,
-      g_param_spec_boolean ("crop", "Crop",
-          "Crop the image to the visible region", THEORA_DEF_CROP,
-          (GParamFlags) G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
-
   if (gst_theora_dec_ctl_is_supported (TH_DECCTL_SET_TELEMETRY_MV)) {
     g_object_class_install_property (gobject_class, PROP_TELEMETRY_MV,
         g_param_spec_int ("visualize-motion-vectors",
@@ -208,7 +201,6 @@ gst_theora_dec_init (GstTheoraDec * dec)
 
   gst_element_add_pad (GST_ELEMENT (dec), dec->srcpad);
 
-  dec->crop = THEORA_DEF_CROP;
   dec->telemetry_mv = THEORA_DEF_TELEMETRY_MV;
   dec->telemetry_mbmode = THEORA_DEF_TELEMETRY_MBMODE;
   dec->telemetry_qi = THEORA_DEF_TELEMETRY_QI;
@@ -776,12 +768,16 @@ theora_handle_comment_packet (GstTheoraDec * dec, ogg_packet * packet)
 }
 
 static GstFlowReturn
-theora_negotiate_pool (GstTheoraDec * dec, GstCaps * caps, GstVideoInfo * info)
+theora_negotiate_pool (GstTheoraDec * dec)
 {
   GstQuery *query;
-  GstBufferPool *pool = NULL;
+  GstBufferPool *pool;
   guint size, min, max, prefix, alignment;
   GstStructure *config;
+  GstCaps *caps;
+
+  /* find the caps of the output buffer */
+  caps = gst_pad_get_current_caps (dec->srcpad);
 
   /* find a pool for the negotiated caps now */
   query = gst_query_new_allocation (caps, TRUE);
@@ -791,35 +787,45 @@ theora_negotiate_pool (GstTheoraDec * dec, GstCaps * caps, GstVideoInfo * info)
     /* we got configuration from our peer, parse them */
     gst_query_parse_allocation_params (query, &size, &min, &max, &prefix,
         &alignment, &pool);
-    size = MAX (size, info->size);
   } else {
     GST_DEBUG_OBJECT (dec, "didn't get downstream ALLOCATION hints");
-    size = info->size;
+    size = 0;
     min = max = 0;
     prefix = 0;
     alignment = 0;
+    pool = NULL;
   }
 
   if (pool == NULL) {
     /* we did not get a pool, make one ourselves then */
-    pool = gst_buffer_pool_new ();
+    pool = gst_video_buffer_pool_new ();
   }
 
   if (dec->pool)
     gst_object_unref (dec->pool);
   dec->pool = pool;
 
+  /* check if downstream supports cropping */
+  dec->has_cropping =
+      gst_query_has_allocation_meta (query, GST_VIDEO_CROP_META_API_TYPE);
+
+  if (dec->has_cropping) {
+    /* we can crop, configure the pool with buffers of caps and size of the
+     * decoded picture size and then crop them with metadata */
+    size = MAX (size, GST_VIDEO_INFO_SIZE (&dec->vinfo));
+    gst_caps_unref (caps);
+    caps = gst_video_info_to_caps (&dec->vinfo);
+  }
+
   config = gst_buffer_pool_get_config (pool);
   gst_buffer_pool_config_set (config, caps, size, min, max, prefix, alignment);
+  gst_caps_unref (caps);
+
   /* just set the option, if the pool can support it we will transparently use
    * it through the video info API. We could also see if the pool support this
    * option and only activate it then. */
   gst_buffer_pool_config_add_option (config, GST_BUFFER_POOL_OPTION_VIDEO_META);
 
-  /* check if downstream supports cropping */
-  dec->has_cropping =
-      gst_query_has_allocation_meta (query, GST_VIDEO_CROP_META_API_TYPE);
-
   GST_DEBUG_OBJECT (dec, "downstream cropping %d", dec->has_cropping);
 
   gst_buffer_pool_set_config (pool, config);
@@ -836,10 +842,10 @@ theora_handle_type_packet (GstTheoraDec * dec, ogg_packet * packet)
 {
   GstCaps *caps;
   GstVideoFormat format;
-  gint width, height;
   gint par_num, par_den;
   GstFlowReturn ret = GST_FLOW_OK;
   GList *walk;
+  GstVideoInfo info;
 
   GST_DEBUG_OBJECT (dec, "fps %d/%d, PAR %d/%d",
       dec->info.fps_numerator, dec->info.fps_denominator,
@@ -891,14 +897,6 @@ theora_handle_type_packet (GstTheoraDec * dec, ogg_packet * packet)
       goto invalid_format;
   }
 
-  if (dec->crop) {
-    width = dec->info.pic_width;
-    height = dec->info.pic_height;
-  } else {
-    /* no cropping, use the encoded dimensions */
-    width = dec->info.frame_width;
-    height = dec->info.frame_height;
-  }
   if (dec->info.pic_width != dec->info.frame_width ||
       dec->info.pic_height != dec->info.frame_height ||
       dec->info.pic_x != 0 || dec->info.pic_y != 0) {
@@ -929,7 +927,9 @@ theora_handle_type_packet (GstTheoraDec * dec, ogg_packet * packet)
     GST_WARNING_OBJECT (dec, "Could not enable BITS mode visualisation");
   }
 
-  gst_video_info_set_format (&dec->vinfo, format, width, height);
+  /* our info contains the dimensions for the coded picture before cropping */
+  gst_video_info_set_format (&dec->vinfo, format, dec->info.frame_width,
+      dec->info.frame_height);
   dec->vinfo.fps_n = dec->info.fps_numerator;
   dec->vinfo.fps_d = dec->info.fps_denominator;
   dec->vinfo.par_n = par_num;
@@ -953,13 +953,16 @@ theora_handle_type_packet (GstTheoraDec * dec, ogg_packet * packet)
       break;
   }
 
-  caps = gst_video_info_to_caps (&dec->vinfo);
+  /* for the output caps we always take the cropped dimensions */
+  info = dec->vinfo;
+  gst_video_info_set_format (&info, format, dec->info.pic_width,
+      dec->info.pic_height);
+  caps = gst_video_info_to_caps (&info);
   gst_pad_set_caps (dec->srcpad, caps);
   gst_caps_unref (caps);
 
-  /* negotiate a bufferpool */
-  if ((ret = theora_negotiate_pool (dec, caps, &dec->vinfo)) != GST_FLOW_OK)
-    goto no_bufferpool;
+  /* make sure we negotiate a bufferpool */
+  gst_pad_mark_reconfigure (dec->srcpad);
 
   dec->have_header = TRUE;
 
@@ -983,10 +986,6 @@ invalid_format:
     GST_ERROR_OBJECT (dec, "Invalid pixel format %d", dec->info.pixel_fmt);
     return GST_FLOW_ERROR;
   }
-no_bufferpool:
-  {
-    return ret;
-  }
 }
 
 static GstFlowReturn
@@ -1113,13 +1112,8 @@ theora_handle_image (GstTheoraDec * dec, th_ycbcr_buffer buf, GstBuffer ** out)
   GstVideoCropMeta *crop;
   gint offset_x, offset_y;
 
-  if (gst_pad_check_reconfigure (dec->srcpad)) {
-    GstCaps *caps;
-
-    caps = gst_pad_get_current_caps (dec->srcpad);
-    theora_negotiate_pool (dec, caps, &dec->vinfo);
-    gst_caps_unref (caps);
-  }
+  if (gst_pad_check_reconfigure (dec->srcpad))
+    theora_negotiate_pool (dec);
 
   result = gst_buffer_pool_acquire_buffer (dec->pool, out, NULL);
   if (G_UNLIKELY (result != GST_FLOW_OK))
@@ -1128,7 +1122,7 @@ theora_handle_image (GstTheoraDec * dec, th_ycbcr_buffer buf, GstBuffer ** out)
   if (!gst_video_frame_map (&frame, &dec->vinfo, *out, GST_MAP_WRITE))
     goto invalid_frame;
 
-  if (dec->crop && !dec->has_cropping) {
+  if (!dec->has_cropping) {
     /* we need to crop the hard way */
     offset_x = dec->info.pic_x;
     offset_y = dec->info.pic_y;
@@ -1650,9 +1644,6 @@ theora_dec_set_property (GObject * object, guint prop_id,
   GstTheoraDec *dec = GST_THEORA_DEC (object);
 
   switch (prop_id) {
-    case PROP_CROP:
-      dec->crop = g_value_get_boolean (value);
-      break;
     case PROP_TELEMETRY_MV:
       dec->telemetry_mv = g_value_get_int (value);
       break;
@@ -1678,9 +1669,6 @@ theora_dec_get_property (GObject * object, guint prop_id,
   GstTheoraDec *dec = GST_THEORA_DEC (object);
 
   switch (prop_id) {
-    case PROP_CROP:
-      g_value_set_boolean (value, dec->crop);
-      break;
     case PROP_TELEMETRY_MV:
       g_value_set_int (value, dec->telemetry_mv);
       break;
diff --git a/ext/theora/gsttheoradec.h b/ext/theora/gsttheoradec.h
index 569fba0a93..e9fe1a8480 100644
--- a/ext/theora/gsttheoradec.h
+++ b/ext/theora/gsttheoradec.h
@@ -82,8 +82,6 @@ struct _GstTheoraDec
   gint telemetry_qi;
   gint telemetry_bits;
 
-  gboolean crop;
-
   /* list of buffers that need timestamps */
   GList *queued;
   /* list of raw output buffers */
diff --git a/gst-libs/gst/video/gstvideofilter.c b/gst-libs/gst/video/gstvideofilter.c
index 97bda075bf..248c482eee 100644
--- a/gst-libs/gst/video/gstvideofilter.c
+++ b/gst-libs/gst/video/gstvideofilter.c
@@ -53,7 +53,7 @@ G_DEFINE_ABSTRACT_TYPE (GstVideoFilter, gst_video_filter,
 /* Answer the allocation query downstream. */
 static gboolean
 gst_video_filter_propose_allocation (GstBaseTransform * trans,
-    gboolean passthrough, GstQuery * query)
+    GstQuery * decide_query, GstQuery * query)
 {
   GstVideoFilter *filter = GST_VIDEO_FILTER_CAST (trans);
   GstVideoInfo info;
@@ -63,9 +63,9 @@ gst_video_filter_propose_allocation (GstBaseTransform * trans,
   guint size;
 
   /* we're passthrough, let the parent implementation hande things */
-  if (passthrough)
+  if (decide_query == NULL)
     return GST_BASE_TRANSFORM_CLASS (parent_class)->propose_allocation (trans,
-        passthrough, query);
+        decide_query, query);
 
   gst_query_parse_allocation (query, &caps, &need_pool);
 
@@ -124,7 +124,8 @@ gst_video_filter_decide_allocation (GstBaseTransform * trans, GstQuery * query)
         GST_BUFFER_POOL_OPTION_VIDEO_META);
     gst_buffer_pool_set_config (pool, config);
   }
-  return TRUE;
+  return GST_BASE_TRANSFORM_CLASS (parent_class)->decide_allocation (trans,
+      query);
 }
 
 
diff --git a/gst/videoconvert/gstvideoconvert.c b/gst/videoconvert/gstvideoconvert.c
index d33779702c..b49239211e 100644
--- a/gst/videoconvert/gstvideoconvert.c
+++ b/gst/videoconvert/gstvideoconvert.c
@@ -49,6 +49,12 @@ GST_DEBUG_CATEGORY (videoconvert_debug);
 #define GST_CAT_DEFAULT videoconvert_debug
 GST_DEBUG_CATEGORY_EXTERN (GST_CAT_PERFORMANCE);
 
+GType gst_video_convert_get_type (void);
+
+static GQuark _colorspace_quark;
+
+#define gst_video_convert_parent_class parent_class
+G_DEFINE_TYPE (GstVideoConvert, gst_video_convert, GST_TYPE_VIDEO_FILTER);
 
 enum
 {
@@ -72,8 +78,6 @@ GST_STATIC_PAD_TEMPLATE ("sink",
     GST_STATIC_CAPS (CSP_VIDEO_CAPS)
     );
 
-GType gst_video_convert_get_type (void);
-
 static void gst_video_convert_set_property (GObject * object,
     guint property_id, const GValue * value, GParamSpec * pspec);
 static void gst_video_convert_get_property (GObject * object,
@@ -305,8 +309,74 @@ invalid_palette:
   }
 }
 
-#define gst_video_convert_parent_class parent_class
-G_DEFINE_TYPE (GstVideoConvert, gst_video_convert, GST_TYPE_VIDEO_FILTER);
+static gboolean
+gst_video_convert_propose_allocation (GstBaseTransform * trans,
+    GstQuery * decide_query, GstQuery * query)
+{
+  gboolean ret;
+  guint i, n_metas;
+
+  /* let parent handle */
+  ret = GST_BASE_TRANSFORM_CLASS (parent_class)->propose_allocation (trans,
+      decide_query, query);
+  /* error or passthrough, we're done */
+  if (!ret || decide_query == NULL)
+    return ret;
+
+  /* non-passthrough, copy all metadata, decide_query does not contain the
+   * metadata anymore that depends on the buffer memory */
+  n_metas = gst_query_get_n_allocation_metas (decide_query);
+  for (i = 0; i < n_metas; i++) {
+    GType api;
+
+    api = gst_query_parse_nth_allocation_meta (decide_query, i);
+    gst_query_add_allocation_meta (query, api);
+  }
+  return ret;
+}
+
+typedef struct
+{
+  GstBaseTransform *trans;
+  GstBuffer *outbuf;
+  GQuark tag;
+} CopyMetaData;
+
+static gboolean
+foreach_metadata (GstBuffer * inbuf, GstMeta ** meta, gpointer user_data)
+{
+  CopyMetaData *data = user_data;
+  const GstMetaInfo *info = (*meta)->info;
+
+  if (info->transform_func) {
+    if (gst_meta_api_type_has_tag (info->api, data->tag)) {
+      /* metadata depends on colorspace. FIXME discard for now until we
+       * have some transform data for it. */
+    } else {
+      GstMetaTransformCopy copy_data = { 0, -1 };
+      /* simply copy then */
+      info->transform_func (data->outbuf, *meta, inbuf,
+          _gst_meta_transform_copy, &copy_data);
+    }
+  }
+  return TRUE;
+}
+
+static gboolean
+gst_video_convert_copy_metadata (GstBaseTransform * trans,
+    GstBuffer * inbuf, GstBuffer * outbuf)
+{
+  CopyMetaData data;
+
+  data.trans = trans;
+  data.outbuf = outbuf;
+  data.tag = _colorspace_quark;
+
+  gst_buffer_foreach_meta (inbuf, foreach_metadata, &data);
+
+  return GST_BASE_TRANSFORM_CLASS (parent_class)->copy_metadata (trans, inbuf,
+      outbuf);
+}
 
 static void
 gst_video_convert_finalize (GObject * obj)
@@ -347,6 +417,10 @@ gst_video_convert_class_init (GstVideoConvertClass * klass)
       GST_DEBUG_FUNCPTR (gst_video_convert_transform_caps);
   gstbasetransform_class->fixate_caps =
       GST_DEBUG_FUNCPTR (gst_video_convert_fixate_caps);
+  gstbasetransform_class->propose_allocation =
+      GST_DEBUG_FUNCPTR (gst_video_convert_propose_allocation);
+  gstbasetransform_class->copy_metadata =
+      GST_DEBUG_FUNCPTR (gst_video_convert_copy_metadata);
 
   gstbasetransform_class->passthrough_on_same_caps = TRUE;
 
@@ -428,6 +502,8 @@ plugin_init (GstPlugin * plugin)
   GST_DEBUG_CATEGORY_INIT (videoconvert_debug, "videoconvert", 0,
       "Colorspace Converter");
 
+  _colorspace_quark = g_quark_from_static_string ("colorspace");
+
   return gst_element_register (plugin, "videoconvert",
       GST_RANK_NONE, GST_TYPE_VIDEO_CONVERT);
 }