diff --git a/docs/libs/gst-plugins-bad-libs-sections.txt b/docs/libs/gst-plugins-bad-libs-sections.txt
index dd6fd030e3..2c3928bd70 100644
--- a/docs/libs/gst-plugins-bad-libs-sections.txt
+++ b/docs/libs/gst-plugins-bad-libs-sections.txt
@@ -771,7 +771,7 @@ GST_GL_COLOR_CONVERT_VIDEO_CAPS
 GstGLColorConvert
 GstGLColorConvertClass
 gst_gl_color_convert_new
-gst_gl_color_convert_set_format
+gst_gl_color_convert_set_caps
 gst_gl_color_convert_perform
 <SUBSECTION Standard>
 GstGLColorConvertPrivate
diff --git a/ext/gl/gstglimagesink.c b/ext/gl/gstglimagesink.c
index 240025c5af..06b9b56c46 100644
--- a/ext/gl/gstglimagesink.c
+++ b/ext/gl/gstglimagesink.c
@@ -682,6 +682,11 @@ gst_glimage_sink_change_state (GstElement * element, GstStateChange transition)
         glimage_sink->upload = NULL;
       }
 
+      if (glimage_sink->convert) {
+        gst_object_unref (glimage_sink->convert);
+        glimage_sink->convert = NULL;
+      }
+
       glimage_sink->window_id = 0;
       //but do not reset glimage_sink->new_window_id
 
@@ -760,6 +765,7 @@ gst_glimage_sink_set_caps (GstBaseSink * bsink, GstCaps * caps)
   GstStructure *structure;
   GstBufferPool *newpool, *oldpool;
   GstCapsFeatures *gl_features;
+  GstCaps *uploaded_caps;
 
   GST_DEBUG ("set caps with %" GST_PTR_FORMAT, caps);
 
@@ -846,14 +852,30 @@ gst_glimage_sink_set_caps (GstBaseSink * bsink, GstCaps * caps)
 
   gl_features =
       gst_caps_features_from_string (GST_CAPS_FEATURE_MEMORY_GL_MEMORY);
+
+  uploaded_caps = gst_caps_copy (caps);
+  gst_caps_set_features (uploaded_caps, 0,
+      gst_caps_features_copy (gl_features));
+  gst_gl_upload_set_caps (glimage_sink->upload, caps, uploaded_caps);
+
   if (glimage_sink->gl_caps)
     gst_caps_unref (glimage_sink->gl_caps);
   glimage_sink->gl_caps = gst_caps_copy (caps);
   gst_caps_set_simple (glimage_sink->gl_caps, "format", G_TYPE_STRING, "RGBA",
       NULL);
-  gst_caps_set_features (glimage_sink->gl_caps, 0, gl_features);
+  gst_caps_set_features (glimage_sink->gl_caps, 0,
+      gst_caps_features_copy (gl_features));
+
+  if (glimage_sink->convert)
+    gst_object_unref (glimage_sink->convert);
+  glimage_sink->convert = gst_gl_color_convert_new (glimage_sink->context);
+  if (!gst_gl_color_convert_set_caps (glimage_sink->convert, uploaded_caps,
+          glimage_sink->gl_caps)) {
+    gst_caps_features_free (gl_features);
+    return FALSE;
+  }
+  gst_caps_features_free (gl_features);
 
-  gst_gl_upload_set_caps (glimage_sink->upload, caps, glimage_sink->gl_caps);
   glimage_sink->caps_change = TRUE;
 
   return TRUE;
@@ -863,7 +885,7 @@ static GstFlowReturn
 gst_glimage_sink_prepare (GstBaseSink * bsink, GstBuffer * buf)
 {
   GstGLImageSink *glimage_sink;
-  GstBuffer *next_buffer = NULL;
+  GstBuffer *uploaded_buffer, *next_buffer = NULL;
   GstVideoFrame gl_frame;
   GstVideoInfo gl_info;
 
@@ -879,17 +901,26 @@ gst_glimage_sink_prepare (GstBaseSink * bsink, GstBuffer * buf)
   if (!_ensure_gl_setup (glimage_sink))
     return GST_FLOW_NOT_NEGOTIATED;
 
-  if (!gst_gl_upload_perform_with_buffer (glimage_sink->upload, buf,
-          &next_buffer))
+  if (gst_gl_upload_perform_with_buffer (glimage_sink->upload, buf,
+          &uploaded_buffer) != GST_GL_UPLOAD_DONE)
     goto upload_failed;
 
+  if (!(next_buffer =
+          gst_gl_color_convert_perform (glimage_sink->convert,
+              uploaded_buffer))) {
+    gst_buffer_unref (uploaded_buffer);
+    goto upload_failed;
+  }
+
   gst_video_info_from_caps (&gl_info, glimage_sink->gl_caps);
 
   if (!gst_video_frame_map (&gl_frame, &gl_info, next_buffer,
           GST_MAP_READ | GST_MAP_GL)) {
+    gst_buffer_unref (uploaded_buffer);
     gst_buffer_unref (next_buffer);
     goto upload_failed;
   }
+  gst_buffer_unref (uploaded_buffer);
 
   glimage_sink->next_tex = *(guint *) gl_frame.data[0];
 
diff --git a/ext/gl/gstglimagesink.h b/ext/gl/gstglimagesink.h
index e99fad9f8f..ab981966a4 100644
--- a/ext/gl/gstglimagesink.h
+++ b/ext/gl/gstglimagesink.h
@@ -69,6 +69,7 @@ struct _GstGLImageSink
     gboolean handle_events;
 
     GstGLUpload *upload;
+    GstGLColorConvert *convert;
     guint      next_tex;
     GstBuffer *next_buffer;
 
diff --git a/ext/gl/gstglmixer.c b/ext/gl/gstglmixer.c
index 68de445806..13a9dead99 100644
--- a/ext/gl/gstglmixer.c
+++ b/ext/gl/gstglmixer.c
@@ -1104,7 +1104,7 @@ gst_gl_mixer_process_textures (GstGLMixer * mix, GstBuffer * outbuf)
     walk = g_list_next (walk);
 
     if (vaggpad->buffer != NULL) {
-      GstBuffer *gl_buf;
+      GstBuffer *uploaded_buf;
       GstCaps *gl_caps;
       GstCapsFeatures *gl_features;
       GstVideoInfo gl_info;
@@ -1119,45 +1119,68 @@ gst_gl_mixer_process_textures (GstGLMixer * mix, GstBuffer * outbuf)
           gst_caps_features_from_string (GST_CAPS_FEATURE_MEMORY_GL_MEMORY);
 
       gl_caps = gst_video_info_to_caps (&gl_info);
-      gst_caps_set_features (gl_caps, 0, gl_features);
+      gst_caps_set_features (gl_caps, 0, gst_caps_features_copy (gl_features));
 
       if (!pad->upload) {
         GstCaps *in_caps = gst_pad_get_current_caps (GST_PAD (pad));
+        GstCaps *upload_caps = gst_caps_copy (in_caps);
 
         pad->upload = gst_gl_upload_new (mix->context);
 
-        gst_gl_upload_set_caps (pad->upload, in_caps, gl_caps);
+        gst_caps_set_features (upload_caps, 0,
+            gst_caps_features_copy (gl_features));
+        gst_gl_upload_set_caps (pad->upload, in_caps, upload_caps);
 
+        if (!pad->convert) {
+          pad->convert = gst_gl_color_convert_new (mix->context);
+
+          gst_gl_color_convert_set_caps (pad->convert, upload_caps, gl_caps);
+        }
+
+        gst_caps_unref (upload_caps);
         gst_caps_unref (in_caps);
       }
 
+      gst_caps_features_free (gl_features);
+      gst_caps_unref (gl_caps);
+
       sync_meta = gst_buffer_get_gl_sync_meta (vaggpad->buffer);
       if (sync_meta)
         gst_gl_sync_meta_wait (sync_meta);
 
-      if (!gst_gl_upload_perform_with_buffer (pad->upload,
-              vaggpad->buffer, &gl_buf)) {
+      if (gst_gl_upload_perform_with_buffer (pad->upload,
+              vaggpad->buffer, &uploaded_buf) != GST_GL_UPLOAD_DONE) {
         ++array_index;
         pad->mapped = FALSE;
-        gst_caps_unref (gl_caps);
         continue;
       }
 
-      if (!gst_video_frame_map (&gl_frame, &gl_info, gl_buf,
+      if (pad->gl_buffer)
+        gst_buffer_unref (pad->gl_buffer);
+
+      if (!(pad->gl_buffer =
+              gst_gl_color_convert_perform (pad->convert, uploaded_buf))) {
+        ++array_index;
+        pad->mapped = FALSE;
+        gst_buffer_unref (uploaded_buf);
+        continue;
+      }
+
+      if (!gst_video_frame_map (&gl_frame, &gl_info, pad->gl_buffer,
               GST_MAP_READ | GST_MAP_GL)) {
         ++array_index;
         pad->mapped = FALSE;
-        gst_buffer_unref (gl_buf);
-        gst_caps_unref (gl_caps);
+        gst_buffer_unref (uploaded_buf);
+        gst_buffer_unref (pad->gl_buffer);
+        pad->gl_buffer = NULL;
         continue;
       }
       pad->mapped = TRUE;
 
       frame->texture = *(guint *) gl_frame.data[0];
 
-      gst_caps_unref (gl_caps);
+      gst_buffer_unref (uploaded_buf);
       gst_video_frame_unmap (&gl_frame);
-      gst_buffer_unref (gl_buf);
     }
     ++array_index;
   }
@@ -1292,11 +1315,21 @@ _clean_upload (GstAggregator * agg, GstAggregatorPad * aggpad, gpointer udata)
 {
   GstGLMixerPad *pad = GST_GL_MIXER_PAD (aggpad);
 
+  if (pad->gl_buffer) {
+    gst_buffer_unref (pad->gl_buffer);
+    pad->gl_buffer = NULL;
+  }
+
   if (pad->upload) {
     gst_object_unref (pad->upload);
     pad->upload = NULL;
   }
 
+  if (pad->convert) {
+    gst_object_unref (pad->convert);
+    pad->convert = NULL;
+  }
+
   return TRUE;
 }
 
diff --git a/ext/gl/gstglmixerpad.h b/ext/gl/gstglmixerpad.h
index d2fb8290c3..83ef7276d4 100644
--- a/ext/gl/gstglmixerpad.h
+++ b/ext/gl/gstglmixerpad.h
@@ -49,7 +49,8 @@ struct _GstGLMixerPad
 
   /* <private> */
   GstGLUpload *upload;
-  guint in_tex_id;
+  GstGLColorConvert *convert;
+  GstBuffer *gl_buffer;
   gboolean mapped;
 };
 
diff --git a/gst-libs/gst/gl/gstglcolorconvert.c b/gst-libs/gst/gl/gstglcolorconvert.c
index cfc12ef080..e448431554 100644
--- a/gst-libs/gst/gl/gstglcolorconvert.c
+++ b/gst-libs/gst/gl/gstglcolorconvert.c
@@ -545,47 +545,76 @@ gst_gl_color_convert_reset (GstGLColorConvert * convert)
   }
 }
 
-static void
-_gst_gl_color_convert_set_format_unlocked (GstGLColorConvert * convert,
-    GstVideoInfo * in_info, GstVideoInfo * out_info)
+static gboolean
+_gst_gl_color_convert_set_caps_unlocked (GstGLColorConvert * convert,
+    GstCaps * in_caps, GstCaps * out_caps)
 {
-  g_return_if_fail (convert != NULL);
-  g_return_if_fail (in_info);
-  g_return_if_fail (out_info);
-  g_return_if_fail (GST_VIDEO_INFO_FORMAT (in_info) !=
-      GST_VIDEO_FORMAT_UNKNOWN);
-  g_return_if_fail (GST_VIDEO_INFO_FORMAT (in_info) !=
-      GST_VIDEO_FORMAT_ENCODED);
-  g_return_if_fail (GST_VIDEO_INFO_FORMAT (out_info) !=
-      GST_VIDEO_FORMAT_UNKNOWN);
-  g_return_if_fail (GST_VIDEO_INFO_FORMAT (out_info) !=
-      GST_VIDEO_FORMAT_ENCODED);
+  GstVideoInfo in_info, out_info;
+  GstCapsFeatures *in_features, *out_features, *gl_features;
 
-  if (gst_video_info_is_equal (&convert->in_info, in_info) &&
-      gst_video_info_is_equal (&convert->out_info, out_info))
-    return;
+  g_return_val_if_fail (convert != NULL, FALSE);
+  g_return_val_if_fail (in_caps, FALSE);
+  g_return_val_if_fail (out_caps, FALSE);
+
+  if (!gst_video_info_from_caps (&in_info, in_caps))
+    g_assert_not_reached ();
+
+  if (!gst_video_info_from_caps (&out_info, out_caps))
+    g_assert_not_reached ();
+
+  g_return_val_if_fail (GST_VIDEO_INFO_FORMAT (&in_info) !=
+      GST_VIDEO_FORMAT_UNKNOWN, FALSE);
+  g_return_val_if_fail (GST_VIDEO_INFO_FORMAT (&in_info) !=
+      GST_VIDEO_FORMAT_ENCODED, FALSE);
+  g_return_val_if_fail (GST_VIDEO_INFO_FORMAT (&out_info) !=
+      GST_VIDEO_FORMAT_UNKNOWN, FALSE);
+  g_return_val_if_fail (GST_VIDEO_INFO_FORMAT (&out_info) !=
+      GST_VIDEO_FORMAT_ENCODED, FALSE);
+
+  gl_features =
+      gst_caps_features_from_string (GST_CAPS_FEATURE_MEMORY_GL_MEMORY);
+  in_features = gst_caps_get_features (in_caps, 0);
+  out_features = gst_caps_get_features (out_caps, 0);
+
+  if (!gst_caps_features_is_equal (in_features, gl_features)
+      || !gst_caps_features_is_equal (out_features, gl_features)) {
+    gst_caps_features_free (gl_features);
+    return FALSE;
+  }
+
+  gst_caps_features_free (gl_features);
+
+  if (gst_video_info_is_equal (&convert->in_info, &in_info) &&
+      gst_video_info_is_equal (&convert->out_info, &out_info))
+    return TRUE;
 
   gst_gl_color_convert_reset (convert);
-  convert->in_info = *in_info;
-  convert->out_info = *out_info;
+  convert->in_info = in_info;
+  convert->out_info = out_info;
   convert->initted = FALSE;
+
+  return TRUE;
 }
 
 /**
- * gst_gl_color_convert_set_format:
+ * gst_gl_color_convert_set_caps:
  * @convert: a #GstGLColorConvert
  * @in_info: input #GstVideoInfo
  * @out_info: output #GstVideoInfo
  *
  * Initializes @convert with the information required for conversion.
  */
-void
-gst_gl_color_convert_set_format (GstGLColorConvert * convert,
-    GstVideoInfo * in_info, GstVideoInfo * out_info)
+gboolean
+gst_gl_color_convert_set_caps (GstGLColorConvert * convert,
+    GstCaps * in_caps, GstCaps * out_caps)
 {
+  gboolean ret;
+
   GST_OBJECT_LOCK (convert);
-  _gst_gl_color_convert_set_format_unlocked (convert, in_info, out_info);
+  ret = _gst_gl_color_convert_set_caps_unlocked (convert, in_caps, out_caps);
   GST_OBJECT_UNLOCK (convert);
+
+  return ret;
 }
 
 /**
@@ -594,7 +623,7 @@ gst_gl_color_convert_set_format (GstGLColorConvert * convert,
  * @inbuf: the texture ids for input formatted according to in_info
  *
  * Converts the data contained by @inbuf using the formats specified by the
- * #GstVideoInfo<!--  -->s passed to gst_gl_color_convert_set_format() 
+ * #GstVideoInfo<!--  -->s passed to gst_gl_color_convert_set_caps() 
  *
  * Returns: a converted #GstBuffer or %NULL%
  */
diff --git a/gst-libs/gst/gl/gstglcolorconvert.h b/gst-libs/gst/gl/gstglcolorconvert.h
index 750064c48c..c7c9caba69 100644
--- a/gst-libs/gst/gl/gstglcolorconvert.h
+++ b/gst-libs/gst/gl/gstglcolorconvert.h
@@ -97,9 +97,9 @@ struct _GstGLColorConvertClass
 
 GstGLColorConvert * gst_gl_color_convert_new (GstGLContext * context);
 
-void     gst_gl_color_convert_set_format    (GstGLColorConvert * convert,
-                                             GstVideoInfo * in_info,
-                                             GstVideoInfo * out_info);
+gboolean    gst_gl_color_convert_set_caps    (GstGLColorConvert * convert,
+                                              GstCaps           * in_caps,
+                                              GstCaps           * out_caps);
 
 GstBuffer * gst_gl_color_convert_perform    (GstGLColorConvert * convert, GstBuffer * inbuf);
 
diff --git a/gst-libs/gst/gl/gstgldownload.c b/gst-libs/gst/gl/gstgldownload.c
index 4e0f412b25..0acc0f3657 100644
--- a/gst-libs/gst/gl/gstgldownload.c
+++ b/gst-libs/gst/gl/gstgldownload.c
@@ -246,6 +246,9 @@ _init_download (GstGLDownload * download)
   GstVideoFormat v_format;
   guint out_width, out_height;
   GstVideoInfo in_info;
+  GstCaps *in_caps, *out_caps;
+  GstCapsFeatures *in_gl_features, *out_gl_features;
+  gboolean res;
 
   v_format = GST_VIDEO_INFO_FORMAT (&download->info);
   out_width = GST_VIDEO_INFO_WIDTH (&download->info);
@@ -266,13 +269,24 @@ _init_download (GstGLDownload * download)
     }
   }
 
+  in_gl_features =
+      gst_caps_features_from_string (GST_CAPS_FEATURE_MEMORY_GL_MEMORY);
   gst_video_info_set_format (&in_info, GST_VIDEO_FORMAT_RGBA, out_width,
       out_height);
+  in_caps = gst_video_info_to_caps (&in_info);
+  gst_caps_set_features (in_caps, 0, in_gl_features);
 
-  gst_gl_color_convert_set_format (download->convert, &in_info,
-      &download->info);
+  out_gl_features =
+      gst_caps_features_from_string (GST_CAPS_FEATURE_MEMORY_GL_MEMORY);
+  out_caps = gst_video_info_to_caps (&download->info);
+  gst_caps_set_features (out_caps, 0, out_gl_features);
 
-  return TRUE;
+  res = gst_gl_color_convert_set_caps (download->convert, in_caps, out_caps);
+
+  gst_caps_unref (in_caps);
+  gst_caps_unref (out_caps);
+
+  return res;
 }
 
 static gboolean
diff --git a/gst-libs/gst/gl/gstglfilter.c b/gst-libs/gst/gl/gstglfilter.c
index adc01b18d7..cf744aa8ee 100644
--- a/gst-libs/gst/gl/gstglfilter.c
+++ b/gst-libs/gst/gl/gstglfilter.c
@@ -253,6 +253,11 @@ gst_gl_filter_reset (GstGLFilter * filter)
     filter->upload = NULL;
   }
 
+  if (filter->in_convert) {
+    gst_object_unref (filter->in_convert);
+    filter->in_convert = NULL;
+  }
+
   if (filter->download) {
     gst_object_unref (filter->download);
     filter->download = NULL;
@@ -309,9 +314,9 @@ gst_gl_filter_reset (GstGLFilter * filter)
     gst_object_unref (filter->context);
   filter->context = NULL;
 
-  if (filter->in_gl_caps) {
-    gst_caps_unref (filter->in_gl_caps);
-    filter->in_gl_caps = NULL;
+  if (filter->in_converted_caps) {
+    gst_caps_unref (filter->in_converted_caps);
+    filter->in_converted_caps = NULL;
   }
 }
 
@@ -1225,26 +1230,34 @@ gst_gl_filter_filter_texture (GstGLFilter * filter, GstBuffer * inbuf,
       gst_caps_features_is_equal (GST_CAPS_FEATURES_MEMORY_SYSTEM_MEMORY,
       gst_caps_get_features (filter->out_caps, 0));
   GstMapFlags out_map_flags = GST_MAP_WRITE;
+  GstBuffer *upload_buffer;
 
   filter_class = GST_GL_FILTER_GET_CLASS (filter);
 
+  if (gst_gl_upload_perform_with_buffer (filter->upload, inbuf,
+          &upload_buffer) != GST_GL_UPLOAD_DONE) {
+    return FALSE;
+  }
+
   if (filter->uploaded_buffer) {
     gst_buffer_unref (filter->uploaded_buffer);
     filter->uploaded_buffer = NULL;
   }
 
-  if (!gst_gl_upload_perform_with_buffer (filter->upload, inbuf,
-          &filter->uploaded_buffer)) {
+  filter->uploaded_buffer =
+      gst_gl_color_convert_perform (filter->in_convert, upload_buffer);
+
+  if (!filter->uploaded_buffer) {
     ret = FALSE;
-    goto upload_error;
+    goto inbuf_error;
   }
 
-  gst_video_info_from_caps (&gl_info, filter->in_gl_caps);
+  gst_video_info_from_caps (&gl_info, filter->in_converted_caps);
 
   if (!gst_video_frame_map (&gl_frame, &gl_info, filter->uploaded_buffer,
           GST_MAP_READ | GST_MAP_GL)) {
     ret = FALSE;
-    goto upload_error;
+    goto inbuf_error;
   }
 
   in_tex = *(guint *) gl_frame.data[0];
@@ -1257,7 +1270,7 @@ gst_gl_filter_filter_texture (GstGLFilter * filter, GstBuffer * inbuf,
   if (!gst_video_frame_map (&out_frame, &filter->out_info, outbuf,
           out_map_flags)) {
     ret = FALSE;
-    goto inbuf_error;
+    goto unmap_out_error;
   }
 
   if (!to_download) {
@@ -1290,10 +1303,11 @@ gst_gl_filter_filter_texture (GstGLFilter * filter, GstBuffer * inbuf,
   }
 
   gst_video_frame_unmap (&gl_frame);
-upload_error:
+unmap_out_error:
   gst_video_frame_unmap (&out_frame);
 inbuf_error:
   gst_gl_upload_release_buffer (filter->upload);
+  gst_buffer_unref (upload_buffer);
 
   return ret;
 }
@@ -1315,6 +1329,7 @@ gst_gl_filter_transform (GstBaseTransform * bt, GstBuffer * inbuf,
   if (!filter->upload) {
     GstCaps *in_caps =
         gst_pad_get_current_caps (GST_BASE_TRANSFORM_SINK_PAD (bt));
+    GstCaps *uploaded_caps;
     GstCapsFeatures *out_features;
     GstVideoInfo out_info;
 
@@ -1325,16 +1340,42 @@ gst_gl_filter_transform (GstBaseTransform * bt, GstBuffer * inbuf,
     out_features =
         gst_caps_features_from_string (GST_CAPS_FEATURE_MEMORY_GL_MEMORY);
 
-    if (filter->in_gl_caps)
-      gst_caps_unref (filter->in_gl_caps);
-    filter->in_gl_caps = gst_video_info_to_caps (&out_info);
-    gst_caps_set_features (filter->in_gl_caps, 0, out_features);
+    uploaded_caps = gst_caps_copy (in_caps);
+    gst_caps_set_features (uploaded_caps, 0, out_features);
 
     filter->upload = gst_gl_upload_new (filter->context);
-    if (!gst_gl_upload_set_caps (filter->upload, in_caps, filter->in_gl_caps))
+    if (!gst_gl_upload_set_caps (filter->upload, in_caps, uploaded_caps)) {
+      gst_caps_unref (in_caps);
       return GST_FLOW_ERROR;
+    }
 
     gst_caps_unref (in_caps);
+
+    if (!filter->in_convert) {
+      GstCapsFeatures *converted_features;
+      GstVideoInfo converted_info;
+
+      gst_video_info_set_format (&converted_info,
+          GST_VIDEO_FORMAT_RGBA,
+          GST_VIDEO_INFO_WIDTH (&filter->in_info),
+          GST_VIDEO_INFO_HEIGHT (&filter->in_info));
+      converted_features =
+          gst_caps_features_from_string (GST_CAPS_FEATURE_MEMORY_GL_MEMORY);
+
+      if (filter->in_converted_caps)
+        gst_caps_unref (filter->in_converted_caps);
+      filter->in_converted_caps = gst_video_info_to_caps (&converted_info);
+      gst_caps_set_features (filter->in_converted_caps, 0, converted_features);
+
+      filter->in_convert = gst_gl_color_convert_new (filter->context);
+      if (!gst_gl_color_convert_set_caps (filter->in_convert, uploaded_caps,
+              filter->in_converted_caps)) {
+        gst_caps_unref (uploaded_caps);
+        return GST_FLOW_ERROR;
+      }
+    }
+
+    gst_caps_unref (uploaded_caps);
   }
 
   g_assert (filter_class->filter || filter_class->filter_texture);
diff --git a/gst-libs/gst/gl/gstglfilter.h b/gst-libs/gst/gl/gstglfilter.h
index a2e4d4db8e..a376742ec8 100644
--- a/gst-libs/gst/gl/gstglfilter.h
+++ b/gst-libs/gst/gl/gstglfilter.h
@@ -75,6 +75,9 @@ struct _GstGLFilter
   GLuint             depthbuffer;
 
   GstGLUpload       *upload;
+  GstGLColorConvert *in_convert;
+  GstCaps           *in_converted_caps;
+
   GstGLDownload     *download;
 
   GstBuffer         *uploaded_buffer;
@@ -88,8 +91,6 @@ struct _GstGLFilter
   GstGLContext      *context;
   GstGLContext      *other_context;
 
-  GstCaps           *in_gl_caps;
-
   GLuint             vao;
   GLuint             vertex_buffer;
   GLint              draw_attr_position_loc;
diff --git a/gst-libs/gst/gl/gstglupload.c b/gst-libs/gst/gl/gstglupload.c
index 39e064da63..782d447e12 100644
--- a/gst-libs/gst/gl/gstglupload.c
+++ b/gst-libs/gst/gl/gstglupload.c
@@ -68,16 +68,6 @@ struct _GstGLUploadPrivate
   int method_i;
 };
 
-typedef enum
-{
-  GST_GL_UPLOAD_DONE = 1,
-  GST_GL_UPLOAD_CONVERT_GL_MEMORY = 2,
-
-  GST_GL_UPLOAD_ERROR = -1,
-  GST_GL_UPLOAD_UNSUPPORTED = -2,
-  GST_GL_UPLOAD_UNSHARED_GL_CONTEXT = 3,
-} GstGLUploadReturn;
-
 typedef enum
 {
   METHOD_FLAG_CAN_SHARE_CONTEXT = 1,
@@ -165,7 +155,7 @@ _gl_memory_upload_perform (gpointer impl, GstBuffer * buffer,
 
   *outbuf = gst_buffer_ref (buffer);
 
-  return GST_GL_UPLOAD_CONVERT_GL_MEMORY;
+  return GST_GL_UPLOAD_DONE;
 }
 
 static void
@@ -277,7 +267,7 @@ _egl_image_upload_perform (gpointer impl, GstBuffer * buffer,
     gst_gl_buffer_pool_replace_last_buffer (GST_GL_BUFFER_POOL (buffer->pool),
         buffer);
 
-  return GST_GL_UPLOAD_CONVERT_GL_MEMORY;
+  return GST_GL_UPLOAD_DONE;
 }
 
 static void
@@ -421,7 +411,7 @@ _upload_meta_upload_perform (gpointer impl, GstBuffer * buffer,
   if (!upload->result)
     return GST_GL_UPLOAD_ERROR;
 
-  return GST_GL_UPLOAD_CONVERT_GL_MEMORY;
+  return GST_GL_UPLOAD_DONE;
 }
 
 static void
@@ -522,7 +512,7 @@ _raw_data_upload_perform (gpointer impl, GstBuffer * buffer,
         gst_memory_ref ((GstMemory *) raw->in_tex[i]));
   }
 
-  return GST_GL_UPLOAD_CONVERT_GL_MEMORY;
+  return GST_GL_UPLOAD_DONE;
 }
 
 static void
@@ -768,11 +758,11 @@ _upload_find_method (GstGLUpload * upload)
  *
  * Returns: whether the upload was successful
  */
-gboolean
+GstGLUploadReturn
 gst_gl_upload_perform_with_buffer (GstGLUpload * upload, GstBuffer * buffer,
     GstBuffer ** outbuf_ptr)
 {
-  GstGLUploadReturn ret;
+  GstGLUploadReturn ret = GST_GL_UPLOAD_ERROR;
 
   g_return_val_if_fail (GST_IS_GL_UPLOAD (upload), FALSE);
   g_return_val_if_fail (GST_IS_BUFFER (buffer), FALSE);
@@ -806,22 +796,6 @@ restart:
     upload->priv->method = &_raw_data_upload;
     upload->priv->method_impl = upload->priv->method->new (upload);
     goto restart;
-  } else if (ret == GST_GL_UPLOAD_CONVERT_GL_MEMORY) {
-    GstBuffer *outbuf;
-
-    if (!upload->priv->convert)
-      upload->priv->convert = gst_gl_color_convert_new (upload->context);
-    gst_gl_color_convert_set_format (upload->priv->convert,
-        &upload->priv->in_info, &upload->priv->out_info);
-
-    if (!(outbuf = gst_gl_color_convert_perform (upload->priv->convert,
-                upload->priv->outbuf))) {
-      gst_gl_upload_release_buffer_unlocked (upload);
-      NEXT_METHOD;
-    }
-
-    gst_buffer_unref (upload->priv->outbuf);
-    upload->priv->outbuf = outbuf;
   } else if (ret == GST_GL_UPLOAD_DONE) {
     /* we are done */
   } else {
@@ -835,7 +809,7 @@ restart:
 
   GST_OBJECT_UNLOCK (upload);
 
-  return upload->priv->outbuf != NULL;
+  return ret;
 
 #undef NEXT_METHOD
 }
diff --git a/gst-libs/gst/gl/gstglupload.h b/gst-libs/gst/gl/gstglupload.h
index 25860ba09c..39666fce60 100644
--- a/gst-libs/gst/gl/gstglupload.h
+++ b/gst-libs/gst/gl/gstglupload.h
@@ -35,6 +35,22 @@ GType gst_gl_upload_get_type (void);
 #define GST_IS_GL_UPLOAD_CLASS(klass) (G_TYPE_CHECK_CLASS_TYPE((klass),GST_TYPE_GL_UPLOAD))
 #define GST_GL_UPLOAD_CAST(obj) ((GstGLUpload*)(obj))
 
+/**
+ * GstGLUploadReturn:
+ * @GST_GL_UPLOAD_DONE: No further processing required
+ * @GST_GL_UPLOAD_ERROR: An unspecified error occured
+ * @GST_GL_UPLOAD_UNSUPPORTED: The configuration is unsupported.
+ */
+typedef enum
+{
+  GST_GL_UPLOAD_DONE = 1,
+
+  GST_GL_UPLOAD_ERROR = -1,
+  GST_GL_UPLOAD_UNSUPPORTED = -2,
+  /* <private> */
+  GST_GL_UPLOAD_UNSHARED_GL_CONTEXT = -3,
+} GstGLUploadReturn;
+
 /**
  * GstGLUpload
  *
@@ -67,8 +83,8 @@ GstGLUpload * gst_gl_upload_new            (GstGLContext * context);
 gboolean      gst_gl_upload_set_caps      (GstGLUpload * upload, GstCaps * in_caps, GstCaps * out_caps);
 void          gst_gl_upload_get_caps      (GstGLUpload * upload, GstCaps ** in_caps, GstCaps ** out_caps);
 
-gboolean gst_gl_upload_perform_with_buffer (GstGLUpload * upload, GstBuffer * buffer, GstBuffer ** outbuf);
-void gst_gl_upload_release_buffer (GstGLUpload * upload);
+GstGLUploadReturn gst_gl_upload_perform_with_buffer (GstGLUpload * upload, GstBuffer * buffer, GstBuffer ** outbuf);
+void              gst_gl_upload_release_buffer (GstGLUpload * upload);
 
 G_END_DECLS
 
diff --git a/tests/check/libs/gstglcolorconvert.c b/tests/check/libs/gstglcolorconvert.c
index 9f87fcced8..0a0b59b7dc 100644
--- a/tests/check/libs/gstglcolorconvert.c
+++ b/tests/check/libs/gstglcolorconvert.c
@@ -120,8 +120,12 @@ check_conversion (TestFrame * frames, guint size)
     gchar *in_data[GST_VIDEO_MAX_PLANES] = { 0 };
     GstGLMemory *in_mem[GST_VIDEO_MAX_PLANES] = { 0 };
     GstVideoFrame in_frame;
+    GstCaps *in_caps;
 
     gst_video_info_set_format (&in_info, in_v_format, in_width, in_height);
+    in_caps = gst_video_info_to_caps (&in_info);
+    gst_caps_set_features (in_caps, 0,
+        gst_caps_features_from_string (GST_CAPS_FEATURE_MEMORY_GL_MEMORY));
 
     for (j = 0; j < GST_VIDEO_INFO_N_PLANES (&in_info); j++) {
       in_data[j] = frames[i].data[j];
@@ -155,15 +159,19 @@ check_conversion (TestFrame * frames, guint size)
       GstVideoFormat out_v_format = frames[j].v_format;
       gchar *out_data[GST_VIDEO_MAX_PLANES] = { 0 };
       GstVideoFrame out_frame;
+      GstCaps *out_caps;
 
       gst_video_info_set_format (&out_info, out_v_format, out_width,
           out_height);
+      out_caps = gst_video_info_to_caps (&out_info);
+      gst_caps_set_features (out_caps, 0,
+          gst_caps_features_from_string (GST_CAPS_FEATURE_MEMORY_GL_MEMORY));
 
       for (k = 0; k < GST_VIDEO_INFO_N_PLANES (&out_info); k++) {
         out_data[k] = frames[j].data[k];
       }
 
-      gst_gl_color_convert_set_format (convert, &in_info, &out_info);
+      gst_gl_color_convert_set_caps (convert, in_caps, out_caps);
 
       /* convert the data */
       outbuf = gst_gl_color_convert_perform (convert, inbuf);
@@ -190,6 +198,7 @@ check_conversion (TestFrame * frames, guint size)
       gst_buffer_unref (outbuf);
     }
 
+    gst_caps_unref (in_caps);
     gst_video_frame_unmap (&in_frame);
     gst_buffer_unref (inbuf);
   }