Merge branch 'master' into 0.11

Conflicts:
	ext/speex/gstspeexdec.c
	ext/speex/gstspeexenc.c
	gst/isomp4/atoms.c
	gst/isomp4/gstqtmux.c

5 files changed, 328 insertions, 1167 deletions

diff --git a/ext/speex/Makefile.am b/ext/speex/Makefile.am
index f2efb3614..fe5523796 100644
--- a/ext/speex/Makefile.am
+++ b/ext/speex/Makefile.am
@@ -1,13 +1,14 @@
 plugin_LTLIBRARIES = libgstspeex.la
 
 libgstspeex_la_SOURCES = gstspeex.c gstspeexdec.c gstspeexenc.c
-libgstspeex_la_CFLAGS = \
+libgstspeex_la_CFLAGS = -DGST_USE_UNSTABLE_API \
 	$(GST_PLUGINS_BASE_CFLAGS) \
 	$(GST_BASE_CFLAGS) \
 	$(GST_CFLAGS) \
 	$(SPEEX_CFLAGS)
 libgstspeex_la_LIBADD = \
-	$(GST_PLUGINS_BASE_LIBS) -lgsttag-$(GST_MAJORMINOR) \
+	$(GST_PLUGINS_BASE_LIBS) \
+	-lgsttag-$(GST_MAJORMINOR) -lgstaudio-$(GST_MAJORMINOR) \
 	$(GST_BASE_LIBS) \
 	$(GST_LIBS) \
 	$(SPEEX_LIBS)
diff --git a/ext/speex/gstspeexdec.c b/ext/speex/gstspeexdec.c
index c5add8268..de4121436 100644
--- a/ext/speex/gstspeexdec.c
+++ b/ext/speex/gstspeexdec.c
@@ -78,53 +78,43 @@ GST_STATIC_PAD_TEMPLATE ("sink",
     );
 
 #define gst_speex_dec_parent_class parent_class
-G_DEFINE_TYPE (GstSpeexDec, gst_speex_dec, GST_TYPE_ELEMENT);
-
-static gboolean speex_dec_sink_event (GstPad * pad, GstEvent * event);
-static GstFlowReturn speex_dec_chain (GstPad * pad, GstBuffer * buf);
-static GstStateChangeReturn speex_dec_change_state (GstElement * element,
-    GstStateChange transition);
-
-static gboolean speex_dec_src_event (GstPad * pad, GstEvent * event);
-static gboolean speex_dec_src_query (GstPad * pad, GstQuery * query);
-static gboolean speex_dec_sink_query (GstPad * pad, GstQuery * query);
-static const GstQueryType *speex_get_src_query_types (GstPad * pad);
-static const GstQueryType *speex_get_sink_query_types (GstPad * pad);
-static gboolean speex_dec_convert (GstPad * pad,
-    GstFormat src_format, gint64 src_value,
-    GstFormat * dest_format, gint64 * dest_value);
+G_DEFINE_TYPE (GstSpeexDec, gst_speex_dec, GST_TYPE_AUDIO_DECODER);
+
+static gboolean gst_speex_dec_start (GstAudioDecoder * dec);
+static gboolean gst_speex_dec_stop (GstAudioDecoder * dec);
+static gboolean gst_speex_dec_set_format (GstAudioDecoder * bdec,
+    GstCaps * caps);
+static GstFlowReturn gst_speex_dec_handle_frame (GstAudioDecoder * dec,
+    GstBuffer * buffer);
 
 static void gst_speex_dec_get_property (GObject * object, guint prop_id,
     GValue * value, GParamSpec * pspec);
 static void gst_speex_dec_set_property (GObject * object, guint prop_id,
     const GValue * value, GParamSpec * pspec);
 
-static GstFlowReturn speex_dec_chain_parse_data (GstSpeexDec * dec,
-    GstBuffer * buf, GstClockTime timestamp, GstClockTime duration);
-
-static GstFlowReturn speex_dec_chain_parse_header (GstSpeexDec * dec,
-    GstBuffer * buf);
-static GstFlowReturn speex_dec_chain_parse_comments (GstSpeexDec * dec,
-    GstBuffer * buf);
-
 static void
 gst_speex_dec_class_init (GstSpeexDecClass * klass)
 {
   GObjectClass *gobject_class;
   GstElementClass *gstelement_class;
+  GstAudioDecoderClass *base_class;
 
   gobject_class = (GObjectClass *) klass;
   gstelement_class = (GstElementClass *) klass;
+  base_class = (GstAudioDecoderClass *) klass;
 
   gobject_class->set_property = gst_speex_dec_set_property;
   gobject_class->get_property = gst_speex_dec_get_property;
 
+  base_class->start = GST_DEBUG_FUNCPTR (gst_speex_dec_start);
+  base_class->stop = GST_DEBUG_FUNCPTR (gst_speex_dec_stop);
+  base_class->set_format = GST_DEBUG_FUNCPTR (gst_speex_dec_set_format);
+  base_class->handle_frame = GST_DEBUG_FUNCPTR (gst_speex_dec_handle_frame);
+
   g_object_class_install_property (G_OBJECT_CLASS (klass), ARG_ENH,
       g_param_spec_boolean ("enh", "Enh", "Enable perceptual enhancement",
           DEFAULT_ENH, G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
 
-  gstelement_class->change_state = GST_DEBUG_FUNCPTR (speex_dec_change_state);
-
   gst_element_class_add_pad_template (gstelement_class,
       gst_static_pad_template_get (&speex_dec_src_factory));
   gst_element_class_add_pad_template (gstelement_class,
@@ -140,7 +130,6 @@ gst_speex_dec_class_init (GstSpeexDecClass * klass)
 static void
 gst_speex_dec_reset (GstSpeexDec * dec)
 {
-  gst_segment_init (&dec->segment, GST_FORMAT_UNDEFINED);
   dec->packetno = 0;
   dec->frame_size = 0;
   dec->frame_duration = 0;
@@ -166,396 +155,38 @@ gst_speex_dec_reset (GstSpeexDec * dec)
 static void
 gst_speex_dec_init (GstSpeexDec * dec)
 {
-  dec->sinkpad =
-      gst_pad_new_from_static_template (&speex_dec_sink_factory, "sink");
-  gst_pad_set_chain_function (dec->sinkpad,
-      GST_DEBUG_FUNCPTR (speex_dec_chain));
-  gst_pad_set_event_function (dec->sinkpad,
-      GST_DEBUG_FUNCPTR (speex_dec_sink_event));
-  gst_pad_set_query_type_function (dec->sinkpad,
-      GST_DEBUG_FUNCPTR (speex_get_sink_query_types));
-  gst_pad_set_query_function (dec->sinkpad,
-      GST_DEBUG_FUNCPTR (speex_dec_sink_query));
-  gst_element_add_pad (GST_ELEMENT (dec), dec->sinkpad);
-
-  dec->srcpad =
-      gst_pad_new_from_static_template (&speex_dec_src_factory, "src");
-  gst_pad_use_fixed_caps (dec->srcpad);
-  gst_pad_set_event_function (dec->srcpad,
-      GST_DEBUG_FUNCPTR (speex_dec_src_event));
-  gst_pad_set_query_type_function (dec->srcpad,
-      GST_DEBUG_FUNCPTR (speex_get_src_query_types));
-  gst_pad_set_query_function (dec->srcpad,
-      GST_DEBUG_FUNCPTR (speex_dec_src_query));
-  gst_element_add_pad (GST_ELEMENT (dec), dec->srcpad);
-
   dec->enh = DEFAULT_ENH;
 
   gst_speex_dec_reset (dec);
 }
 
 static gboolean
-speex_dec_sink_setcaps (GstPad * pad, GstCaps * caps)
-{
-  GstSpeexDec *dec = GST_SPEEX_DEC (gst_pad_get_parent (pad));
-  gboolean ret = TRUE;
-  GstStructure *s;
-  const GValue *streamheader;
-
-  s = gst_caps_get_structure (caps, 0);
-  if ((streamheader = gst_structure_get_value (s, "streamheader")) &&
-      G_VALUE_HOLDS (streamheader, GST_TYPE_ARRAY) &&
-      gst_value_array_get_size (streamheader) >= 2) {
-    const GValue *header, *vorbiscomment;
-    GstBuffer *buf;
-    GstFlowReturn res = GST_FLOW_OK;
-
-    header = gst_value_array_get_value (streamheader, 0);
-    if (header && G_VALUE_HOLDS (header, GST_TYPE_BUFFER)) {
-      buf = gst_value_get_buffer (header);
-      res = speex_dec_chain_parse_header (dec, buf);
-      if (res != GST_FLOW_OK)
-        goto done;
-      gst_buffer_replace (&dec->streamheader, buf);
-    }
-
-    vorbiscomment = gst_value_array_get_value (streamheader, 1);
-    if (vorbiscomment && G_VALUE_HOLDS (vorbiscomment, GST_TYPE_BUFFER)) {
-      buf = gst_value_get_buffer (vorbiscomment);
-      res = speex_dec_chain_parse_comments (dec, buf);
-      if (res != GST_FLOW_OK)
-        goto done;
-      gst_buffer_replace (&dec->vorbiscomment, buf);
-    }
-  }
-
-done:
-  gst_object_unref (dec);
-  return ret;
-}
-
-static gboolean
-speex_dec_convert (GstPad * pad,
-    GstFormat src_format, gint64 src_value,
-    GstFormat * dest_format, gint64 * dest_value)
-{
-  gboolean res = TRUE;
-  GstSpeexDec *dec;
-  guint64 scale = 1;
-
-  dec = GST_SPEEX_DEC (gst_pad_get_parent (pad));
-
-  if (src_format == *dest_format) {
-    *dest_value = src_value;
-    res = TRUE;
-    goto cleanup;
-  }
-
-  if (dec->packetno < 1) {
-    res = FALSE;
-    goto cleanup;
-  }
-
-  if (pad == dec->sinkpad &&
-      (src_format == GST_FORMAT_BYTES || *dest_format == GST_FORMAT_BYTES)) {
-    res = FALSE;
-    goto cleanup;
-  }
-
-  switch (src_format) {
-    case GST_FORMAT_TIME:
-      switch (*dest_format) {
-        case GST_FORMAT_BYTES:
-          scale = 2 * dec->header->nb_channels;
-        case GST_FORMAT_DEFAULT:
-          *dest_value =
-              gst_util_uint64_scale_int (scale * src_value, dec->header->rate,
-              GST_SECOND);
-          break;
-        default:
-          res = FALSE;
-          break;
-      }
-      break;
-    case GST_FORMAT_DEFAULT:
-      switch (*dest_format) {
-        case GST_FORMAT_BYTES:
-          *dest_value = src_value * 2 * dec->header->nb_channels;
-          break;
-        case GST_FORMAT_TIME:
-          *dest_value =
-              gst_util_uint64_scale_int (src_value, GST_SECOND,
-              dec->header->rate);
-          break;
-        default:
-          res = FALSE;
-          break;
-      }
-      break;
-    case GST_FORMAT_BYTES:
-      switch (*dest_format) {
-        case GST_FORMAT_DEFAULT:
-          *dest_value = src_value / (2 * dec->header->nb_channels);
-          break;
-        case GST_FORMAT_TIME:
-          *dest_value = gst_util_uint64_scale_int (src_value, GST_SECOND,
-              dec->header->rate * 2 * dec->header->nb_channels);
-          break;
-        default:
-          res = FALSE;
-          break;
-      }
-      break;
-    default:
-      res = FALSE;
-      break;
-  }
-
-cleanup:
-  gst_object_unref (dec);
-  return res;
-}
-
-static const GstQueryType *
-speex_get_sink_query_types (GstPad * pad)
-{
-  static const GstQueryType speex_dec_sink_query_types[] = {
-    GST_QUERY_CONVERT,
-    0
-  };
-
-  return speex_dec_sink_query_types;
-}
-
-static gboolean
-speex_dec_sink_query (GstPad * pad, GstQuery * query)
-{
-  GstSpeexDec *dec;
-  gboolean res;
-
-  dec = GST_SPEEX_DEC (gst_pad_get_parent (pad));
-
-  switch (GST_QUERY_TYPE (query)) {
-    case GST_QUERY_CONVERT:
-    {
-      GstFormat src_fmt, dest_fmt;
-      gint64 src_val, dest_val;
-
-      gst_query_parse_convert (query, &src_fmt, &src_val, &dest_fmt, &dest_val);
-      res = speex_dec_convert (pad, src_fmt, src_val, &dest_fmt, &dest_val);
-      if (res) {
-        gst_query_set_convert (query, src_fmt, src_val, dest_fmt, dest_val);
-      }
-      break;
-    }
-    default:
-      res = gst_pad_query_default (pad, query);
-      break;
-  }
-
-  gst_object_unref (dec);
-  return res;
-}
-
-static const GstQueryType *
-speex_get_src_query_types (GstPad * pad)
-{
-  static const GstQueryType speex_dec_src_query_types[] = {
-    GST_QUERY_POSITION,
-    GST_QUERY_DURATION,
-    0
-  };
-
-  return speex_dec_src_query_types;
-}
-
-static gboolean
-speex_dec_src_query (GstPad * pad, GstQuery * query)
-{
-  GstSpeexDec *dec;
-  gboolean res = FALSE;
-
-  dec = GST_SPEEX_DEC (gst_pad_get_parent (pad));
-
-  /* FIXME: why not just pass position/duration queries upstream to demuxer? */
-  switch (GST_QUERY_TYPE (query)) {
-    case GST_QUERY_POSITION:{
-      GstSegment segment;
-      GstFormat format;
-      gint64 cur;
-
-      gst_query_parse_position (query, &format, NULL);
-
-      GST_PAD_STREAM_LOCK (dec->sinkpad);
-      segment = dec->segment;
-      GST_PAD_STREAM_UNLOCK (dec->sinkpad);
-
-      if (segment.format != GST_FORMAT_TIME) {
-        GST_DEBUG_OBJECT (dec, "segment not initialised yet");
-        break;
-      }
-
-      if ((res = speex_dec_convert (dec->srcpad, GST_FORMAT_TIME,
-                  segment.position, &format, &cur))) {
-        gst_query_set_position (query, format, cur);
-      }
-      break;
-    }
-    case GST_QUERY_DURATION:{
-      GstFormat format;
-      gint64 dur;
-
-      /* get duration from demuxer */
-      if (!gst_pad_query_peer_duration (dec->sinkpad, GST_FORMAT_TIME, &dur))
-        break;
-
-      gst_query_parse_duration (query, &format, NULL);
-
-      /* and convert it into the requested format */
-      if ((res = speex_dec_convert (dec->srcpad, GST_FORMAT_TIME,
-                  dur, &format, &dur))) {
-        gst_query_set_duration (query, format, dur);
-      }
-      break;
-    }
-    default:
-      res = gst_pad_query_default (pad, query);
-      break;
-  }
-
-  gst_object_unref (dec);
-  return res;
-}
-
-static gboolean
-speex_dec_src_event (GstPad * pad, GstEvent * event)
+gst_speex_dec_start (GstAudioDecoder * dec)
 {
-  gboolean res = FALSE;
-  GstSpeexDec *dec = GST_SPEEX_DEC (gst_pad_get_parent (pad));
-
-  GST_LOG_OBJECT (dec, "handling %s event", GST_EVENT_TYPE_NAME (event));
-
-  switch (GST_EVENT_TYPE (event)) {
-    case GST_EVENT_SEEK:{
-      GstFormat format, tformat;
-      gdouble rate;
-      GstEvent *real_seek;
-      GstSeekFlags flags;
-      GstSeekType cur_type, stop_type;
-      gint64 cur, stop;
-      gint64 tcur, tstop;
-
-      gst_event_parse_seek (event, &rate, &format, &flags, &cur_type, &cur,
-          &stop_type, &stop);
-
-      /* we have to ask our peer to seek to time here as we know
-       * nothing about how to generate a granulepos from the src
-       * formats or anything.
-       *
-       * First bring the requested format to time
-       */
-      tformat = GST_FORMAT_TIME;
-      if (!(res = speex_dec_convert (pad, format, cur, &tformat, &tcur)))
-        break;
-      if (!(res = speex_dec_convert (pad, format, stop, &tformat, &tstop)))
-        break;
+  GstSpeexDec *sd = GST_SPEEX_DEC (dec);
 
-      /* then seek with time on the peer */
-      real_seek = gst_event_new_seek (rate, GST_FORMAT_TIME,
-          flags, cur_type, tcur, stop_type, tstop);
+  GST_DEBUG_OBJECT (dec, "start");
+  gst_speex_dec_reset (sd);
 
-      GST_LOG_OBJECT (dec, "seek to %" GST_TIME_FORMAT, GST_TIME_ARGS (tcur));
-
-      res = gst_pad_push_event (dec->sinkpad, real_seek);
-      gst_event_unref (event);
-      break;
-    }
-    default:
-      res = gst_pad_event_default (pad, event);
-      break;
-  }
+  /* we know about concealment */
+  gst_audio_decoder_set_plc_aware (dec, TRUE);
 
-  gst_object_unref (dec);
-  return res;
+  return TRUE;
 }
 
 static gboolean
-speex_dec_sink_event (GstPad * pad, GstEvent * event)
+gst_speex_dec_stop (GstAudioDecoder * dec)
 {
-  GstSpeexDec *dec;
-  gboolean ret = FALSE;
-
-  dec = GST_SPEEX_DEC (gst_pad_get_parent (pad));
-
-  GST_LOG_OBJECT (dec, "handling %s event", GST_EVENT_TYPE_NAME (event));
-
-  switch (GST_EVENT_TYPE (event)) {
-    case GST_EVENT_CAPS:
-    {
-      GstCaps *caps;
-
-      gst_event_parse_caps (event, &caps);
-      ret = speex_dec_sink_setcaps (pad, caps);
-      gst_event_unref (event);
-      break;
-    }
-    case GST_EVENT_SEGMENT:{
-      GstSegment segment;
-
-      gst_event_copy_segment (event, &segment);
-
-      if (segment.format != GST_FORMAT_TIME)
-        goto newseg_wrong_format;
-
-      if (segment.rate <= 0.0)
-        goto newseg_wrong_rate;
-
-#if 0
-      if (update) {
-        /* time progressed without data, see if we can fill the gap with
-         * some concealment data */
-        if (dec->segment.position < start) {
-          GstClockTime duration;
-
-          duration = start - dec->segment.position;
-          speex_dec_chain_parse_data (dec, NULL, dec->segment.position,
-              duration);
-        }
-      }
-#endif
-
-      /* now configure the values */
-      dec->segment = segment;
-
-      GST_DEBUG_OBJECT (dec, "segment now: %" GST_SEGMENT_FORMAT, &segment);
-      ret = gst_pad_push_event (dec->srcpad, event);
-      break;
-    }
-    default:
-      ret = gst_pad_event_default (pad, event);
-      break;
-  }
+  GstSpeexDec *sd = GST_SPEEX_DEC (dec);
 
-  gst_object_unref (dec);
-  return ret;
+  GST_DEBUG_OBJECT (dec, "stop");
+  gst_speex_dec_reset (sd);
 
-  /* ERRORS */
-newseg_wrong_format:
-  {
-    GST_DEBUG_OBJECT (dec, "received non TIME newsegment");
-    gst_object_unref (dec);
-    return FALSE;
-  }
-newseg_wrong_rate:
-  {
-    GST_DEBUG_OBJECT (dec, "negative rates not supported yet");
-    gst_object_unref (dec);
-    return FALSE;
-  }
+  return TRUE;
 }
 
 static GstFlowReturn
-speex_dec_chain_parse_header (GstSpeexDec * dec, GstBuffer * buf)
+gst_speex_dec_parse_header (GstSpeexDec * dec, GstBuffer * buf)
 {
   GstCaps *caps;
   char *data;
@@ -603,7 +234,7 @@ speex_dec_chain_parse_header (GstSpeexDec * dec, GstBuffer * buf)
       "rate", G_TYPE_INT, dec->header->rate,
       "channels", G_TYPE_INT, dec->header->nb_channels, NULL);
 
-  if (!gst_pad_set_caps (dec->srcpad, caps))
+  if (!gst_pad_set_caps (GST_AUDIO_DECODER_SRC_PAD (dec), caps))
     goto nego_failed;
 
   gst_caps_unref (caps);
@@ -640,7 +271,7 @@ nego_failed:
 }
 
 static GstFlowReturn
-speex_dec_chain_parse_comments (GstSpeexDec * dec, GstBuffer * buf)
+gst_speex_dec_parse_comments (GstSpeexDec * dec, GstBuffer * buf)
 {
   GstTagList *list;
   gchar *ver, *encoder = NULL;
@@ -675,7 +306,8 @@ speex_dec_chain_parse_comments (GstSpeexDec * dec, GstBuffer * buf)
 
   GST_INFO_OBJECT (dec, "tags: %" GST_PTR_FORMAT, list);
 
-  gst_element_found_tags_for_pad (GST_ELEMENT (dec), dec->srcpad, list);
+  gst_element_found_tags_for_pad (GST_ELEMENT (dec),
+      GST_AUDIO_DECODER_SRC_PAD (dec), list);
 
   g_free (encoder);
   g_free (ver);
@@ -683,9 +315,47 @@ speex_dec_chain_parse_comments (GstSpeexDec * dec, GstBuffer * buf)
   return GST_FLOW_OK;
 }
 
+static gboolean
+gst_speex_dec_set_format (GstAudioDecoder * bdec, GstCaps * caps)
+{
+  GstSpeexDec *dec = GST_SPEEX_DEC (bdec);
+  gboolean ret = TRUE;
+  GstStructure *s;
+  const GValue *streamheader;
+
+  s = gst_caps_get_structure (caps, 0);
+  if ((streamheader = gst_structure_get_value (s, "streamheader")) &&
+      G_VALUE_HOLDS (streamheader, GST_TYPE_ARRAY) &&
+      gst_value_array_get_size (streamheader) >= 2) {
+    const GValue *header, *vorbiscomment;
+    GstBuffer *buf;
+    GstFlowReturn res = GST_FLOW_OK;
+
+    header = gst_value_array_get_value (streamheader, 0);
+    if (header && G_VALUE_HOLDS (header, GST_TYPE_BUFFER)) {
+      buf = gst_value_get_buffer (header);
+      res = gst_speex_dec_parse_header (dec, buf);
+      if (res != GST_FLOW_OK)
+        goto done;
+      gst_buffer_replace (&dec->streamheader, buf);
+    }
+
+    vorbiscomment = gst_value_array_get_value (streamheader, 1);
+    if (vorbiscomment && G_VALUE_HOLDS (vorbiscomment, GST_TYPE_BUFFER)) {
+      buf = gst_value_get_buffer (vorbiscomment);
+      res = gst_speex_dec_parse_comments (dec, buf);
+      if (res != GST_FLOW_OK)
+        goto done;
+      gst_buffer_replace (&dec->vorbiscomment, buf);
+    }
+  }
+
+done:
+  return ret;
+}
+
 static GstFlowReturn
-speex_dec_chain_parse_data (GstSpeexDec * dec, GstBuffer * buf,
-    GstClockTime timestamp, GstClockTime duration)
+gst_speex_dec_parse_data (GstSpeexDec * dec, GstBuffer * buf)
 {
   GstFlowReturn res = GST_FLOW_OK;
   gint i, fpp;
@@ -696,13 +366,7 @@ speex_dec_chain_parse_data (GstSpeexDec * dec, GstBuffer * buf,
   if (!dec->frame_duration)
     goto not_negotiated;
 
-  if (timestamp != -1) {
-    dec->segment.position = timestamp;
-  } else {
-    timestamp = dec->segment.position;
-  }
-
-  if (buf) {
+  if (G_LIKELY (gst_buffer_get_size (buf))) {
     /* send data to the bitstream */
     data = gst_buffer_map (buf, &size, NULL, GST_MAP_READ);
     speex_bits_read_from (&dec->bits, data, size);
@@ -711,16 +375,16 @@ speex_dec_chain_parse_data (GstSpeexDec * dec, GstBuffer * buf,
     fpp = dec->header->frames_per_packet;
     bits = &dec->bits;
 
-    GST_DEBUG_OBJECT (dec, "received buffer of size %u, fpp %d, %d bits", size,
-        fpp, speex_bits_remaining (bits));
+    GST_DEBUG_OBJECT (dec, "received buffer of size %u, fpp %d, %d bits",
+        size, fpp, speex_bits_remaining (bits));
   } else {
+    /* FIXME ? actually consider how much concealment is needed */
     /* concealment data, pass NULL as the bits parameters */
     GST_DEBUG_OBJECT (dec, "creating concealment data");
     fpp = dec->header->frames_per_packet;
     bits = NULL;
   }
 
-
   /* now decode each frame, catering for unknown number of them (e.g. rtp) */
   for (i = 0; i < fpp; i++) {
     GstBuffer *outbuf;
@@ -730,9 +394,10 @@ speex_dec_chain_parse_data (GstSpeexDec * dec, GstBuffer * buf,
     GST_LOG_OBJECT (dec, "decoding frame %d/%d, %d bits remaining", i, fpp,
         bits ? speex_bits_remaining (bits) : -1);
 #if 0
-    res = gst_pad_alloc_buffer_and_set_caps (dec->srcpad,
+    res =
+        gst_pad_alloc_buffer_and_set_caps (GST_AUDIO_DECODER_SRC_PAD (dec),
         GST_BUFFER_OFFSET_NONE, dec->frame_size * dec->header->nb_channels * 2,
-        GST_PAD_CAPS (dec->srcpad), &outbuf);
+        GST_PAD_CAPS (GST_AUDIO_DECODER_SRC_PAD (dec)), &outbuf);
 
     if (res != GST_FLOW_OK) {
       GST_DEBUG_OBJECT (dec, "buf alloc flow: %s", gst_flow_get_name (res));
@@ -754,39 +419,27 @@ speex_dec_chain_parse_data (GstSpeexDec * dec, GstBuffer * buf,
       if (fpp == 0 && speex_bits_remaining (bits) < 8) {
         /* if we did not know how many frames to expect, then we get this
            at the end if there are leftover bits to pad to the next byte */
+        GST_DEBUG_OBJECT (dec, "Discarding leftover bits");
       } else {
         GST_WARNING_OBJECT (dec, "Unexpected end of stream found");
       }
+      gst_audio_decoder_finish_frame (GST_AUDIO_DECODER (dec), NULL, 1);
       gst_buffer_unref (outbuf);
-      outbuf = NULL;
-      break;
     } else if (ret == -2) {
       GST_WARNING_OBJECT (dec, "Decoding error: corrupted stream?");
+      gst_audio_decoder_finish_frame (GST_AUDIO_DECODER (dec), NULL, 1);
       gst_buffer_unref (outbuf);
-      outbuf = NULL;
-      break;
     }
 
     if (bits && speex_bits_remaining (bits) < 0) {
       GST_WARNING_OBJECT (dec, "Decoding overflow: corrupted stream?");
+      gst_audio_decoder_finish_frame (GST_AUDIO_DECODER (dec), NULL, 1);
       gst_buffer_unref (outbuf);
-      outbuf = NULL;
-      break;
     }
     if (dec->header->nb_channels == 2)
       speex_decode_stereo_int (out_data, dec->frame_size, dec->stereo);
 
-    GST_BUFFER_TIMESTAMP (outbuf) = timestamp;
-    GST_BUFFER_DURATION (outbuf) = dec->frame_duration;
-
-    dec->segment.position += dec->frame_duration;
-    timestamp = dec->segment.position;
-
-    GST_LOG_OBJECT (dec, "pushing buffer with ts=%" GST_TIME_FORMAT ", dur=%"
-        GST_TIME_FORMAT, GST_TIME_ARGS (timestamp),
-        GST_TIME_ARGS (dec->frame_duration));
-
-    res = gst_pad_push (dec->srcpad, outbuf);
+    res = gst_audio_decoder_finish_frame (GST_AUDIO_DECODER (dec), outbuf, 1);
 
     if (res != GST_FLOW_OK) {
       GST_DEBUG_OBJECT (dec, "flow: %s", gst_flow_get_name (res));
@@ -826,48 +479,55 @@ memcmp_buffers (GstBuffer * buf1, GstBuffer * buf2)
 }
 
 static GstFlowReturn
-speex_dec_chain (GstPad * pad, GstBuffer * buf)
+gst_speex_dec_handle_frame (GstAudioDecoder * bdec, GstBuffer * buf)
 {
   GstFlowReturn res;
   GstSpeexDec *dec;
 
-  dec = GST_SPEEX_DEC (gst_pad_get_parent (pad));
+  /* no fancy draining */
+  if (G_UNLIKELY (!buf))
+    return GST_FLOW_OK;
+
+  dec = GST_SPEEX_DEC (bdec);
 
   /* If we have the streamheader and vorbiscomment from the caps already
    * ignore them here */
   if (dec->streamheader && dec->vorbiscomment) {
     if (memcmp_buffers (dec->streamheader, buf)) {
+      GST_DEBUG_OBJECT (dec, "found streamheader");
+      gst_audio_decoder_finish_frame (bdec, NULL, 1);
       res = GST_FLOW_OK;
     } else if (memcmp_buffers (dec->vorbiscomment, buf)) {
+      GST_DEBUG_OBJECT (dec, "found vorbiscomments");
+      gst_audio_decoder_finish_frame (bdec, NULL, 1);
       res = GST_FLOW_OK;
     } else {
-      res =
-          speex_dec_chain_parse_data (dec, buf, GST_BUFFER_TIMESTAMP (buf),
-          GST_BUFFER_DURATION (buf));
+      res = gst_speex_dec_parse_data (dec, buf);
     }
   } else {
     /* Otherwise fall back to packet counting and assume that the
      * first two packets are the headers. */
     switch (dec->packetno) {
       case 0:
-        res = speex_dec_chain_parse_header (dec, buf);
+        GST_DEBUG_OBJECT (dec, "counted streamheader");
+        res = gst_speex_dec_parse_header (dec, buf);
+        gst_audio_decoder_finish_frame (bdec, NULL, 1);
         break;
       case 1:
-        res = speex_dec_chain_parse_comments (dec, buf);
+        GST_DEBUG_OBJECT (dec, "counted vorbiscomments");
+        res = gst_speex_dec_parse_comments (dec, buf);
+        gst_audio_decoder_finish_frame (bdec, NULL, 1);
         break;
       default:
-        res =
-            speex_dec_chain_parse_data (dec, buf, GST_BUFFER_TIMESTAMP (buf),
-            GST_BUFFER_DURATION (buf));
+      {
+        res = gst_speex_dec_parse_data (dec, buf);
         break;
+      }
     }
   }
 
   dec->packetno++;
 
-  gst_buffer_unref (buf);
-  gst_object_unref (dec);
-
   return res;
 }
 
@@ -906,37 +566,3 @@ gst_speex_dec_set_property (GObject * object, guint prop_id,
       break;
   }
 }
-
-
-static GstStateChangeReturn
-speex_dec_change_state (GstElement * element, GstStateChange transition)
-{
-  GstStateChangeReturn ret;
-  GstSpeexDec *dec = GST_SPEEX_DEC (element);
-
-  switch (transition) {
-    case GST_STATE_CHANGE_NULL_TO_READY:
-    case GST_STATE_CHANGE_READY_TO_PAUSED:
-    case GST_STATE_CHANGE_PAUSED_TO_PLAYING:
-    default:
-      break;
-  }
-
-  ret = GST_ELEMENT_CLASS (parent_class)->change_state (element, transition);
-  if (ret != GST_STATE_CHANGE_SUCCESS)
-    return ret;
-
-  switch (transition) {
-    case GST_STATE_CHANGE_PLAYING_TO_PAUSED:
-      break;
-    case GST_STATE_CHANGE_PAUSED_TO_READY:
-      gst_speex_dec_reset (dec);
-      break;
-    case GST_STATE_CHANGE_READY_TO_NULL:
-      break;
-    default:
-      break;
-  }
-
-  return ret;
-}
diff --git a/ext/speex/gstspeexdec.h b/ext/speex/gstspeexdec.h
index 660d8053d..8187af87c 100644
--- a/ext/speex/gstspeexdec.h
+++ b/ext/speex/gstspeexdec.h
@@ -22,6 +22,8 @@
 #define __GST_SPEEX_DEC_H__
 
 #include <gst/gst.h>
+#include <gst/audio/gstaudiodecoder.h>
+
 #include <speex/speex.h>
 #include <speex/speex_callbacks.h>
 #include <speex/speex_header.h>
@@ -44,11 +46,7 @@ typedef struct _GstSpeexDec GstSpeexDec;
 typedef struct _GstSpeexDecClass GstSpeexDecClass;
 
 struct _GstSpeexDec {
-  GstElement            element;
-
-  /* pads */
-  GstPad                *sinkpad;
-  GstPad                *srcpad;
+  GstAudioDecoder   element;
 
   void                  *state;
   SpeexStereoState      *stereo;
@@ -67,14 +65,12 @@ struct _GstSpeexDec {
   GstClockTime          frame_duration;
   guint64               packetno;
 
-  GstSegment            segment;    /* STREAM LOCK */
-
   GstBuffer             *streamheader;
   GstBuffer             *vorbiscomment;
 };
 
 struct _GstSpeexDecClass {
-  GstElementClass parent_class;
+  GstAudioDecoderClass parent_class;
 };
 
 GType gst_speex_dec_get_type (void);
diff --git a/ext/speex/gstspeexenc.c b/ext/speex/gstspeexenc.c
index e202d9224..e4eda25cf 100644
--- a/ext/speex/gstspeexenc.c
+++ b/ext/speex/gstspeexenc.c
@@ -113,43 +113,28 @@ gst_speex_enc_mode_get_type (void)
   return speex_enc_mode_type;
 }
 
-#if 0
-static const GstFormat *
-gst_speex_enc_get_formats (GstPad * pad)
-{
-  static const GstFormat src_formats[] = {
-    GST_FORMAT_BYTES,
-    GST_FORMAT_TIME,
-    0
-  };
-  static const GstFormat sink_formats[] = {
-    GST_FORMAT_BYTES,
-    GST_FORMAT_DEFAULT,
-    GST_FORMAT_TIME,
-    0
-  };
-
-  return (GST_PAD_IS_SRC (pad) ? src_formats : sink_formats);
-}
-#endif
-
 static void gst_speex_enc_finalize (GObject * object);
 
-static gboolean gst_speex_enc_sink_event (GstPad * pad, GstEvent * event);
-static GstFlowReturn gst_speex_enc_chain (GstPad * pad, GstBuffer * buf);
 static gboolean gst_speex_enc_setup (GstSpeexEnc * enc);
 
 static void gst_speex_enc_get_property (GObject * object, guint prop_id,
     GValue * value, GParamSpec * pspec);
 static void gst_speex_enc_set_property (GObject * object, guint prop_id,
     const GValue * value, GParamSpec * pspec);
-static GstStateChangeReturn gst_speex_enc_change_state (GstElement * element,
-    GstStateChange transition);
 
-static GstFlowReturn gst_speex_enc_encode (GstSpeexEnc * enc, gboolean flush);
+static GstFlowReturn gst_speex_enc_encode (GstSpeexEnc * enc, GstBuffer * buf);
+
+static gboolean gst_speex_enc_start (GstAudioEncoder * enc);
+static gboolean gst_speex_enc_stop (GstAudioEncoder * enc);
+static gboolean gst_speex_enc_set_format (GstAudioEncoder * enc,
+    GstAudioInfo * info);
+static GstFlowReturn gst_speex_enc_handle_frame (GstAudioEncoder * enc,
+    GstBuffer * in_buf);
+static gboolean gst_speex_enc_sink_event (GstAudioEncoder * enc,
+    GstEvent * event);
 
 #define gst_speex_enc_parent_class parent_class
-G_DEFINE_TYPE_WITH_CODE (GstSpeexEnc, gst_speex_enc, GST_TYPE_ELEMENT,
+G_DEFINE_TYPE_WITH_CODE (GstSpeexEnc, gst_speex_enc, GST_TYPE_AUDIO_ENCODER,
     G_IMPLEMENT_INTERFACE (GST_TYPE_TAG_SETTER, NULL);
     G_IMPLEMENT_INTERFACE (GST_TYPE_PRESET, NULL));
 
@@ -158,62 +143,67 @@ gst_speex_enc_class_init (GstSpeexEncClass * klass)
 {
   GObjectClass *gobject_class;
   GstElementClass *gstelement_class;
+  GstAudioEncoderClass *base_class;
 
   gobject_class = (GObjectClass *) klass;
   gstelement_class = (GstElementClass *) klass;
+  base_class = (GstAudioEncoderClass *) klass;
 
   gobject_class->finalize = gst_speex_enc_finalize;
   gobject_class->set_property = gst_speex_enc_set_property;
   gobject_class->get_property = gst_speex_enc_get_property;
 
+  base_class->start = GST_DEBUG_FUNCPTR (gst_speex_enc_start);
+  base_class->stop = GST_DEBUG_FUNCPTR (gst_speex_enc_stop);
+  base_class->set_format = GST_DEBUG_FUNCPTR (gst_speex_enc_set_format);
+  base_class->handle_frame = GST_DEBUG_FUNCPTR (gst_speex_enc_handle_frame);
+  base_class->event = GST_DEBUG_FUNCPTR (gst_speex_enc_sink_event);
+
   g_object_class_install_property (G_OBJECT_CLASS (klass), PROP_QUALITY,
       g_param_spec_float ("quality", "Quality", "Encoding quality",
           0.0, 10.0, DEFAULT_QUALITY,
-          G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
+          G_PARAM_CONSTRUCT | G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
   g_object_class_install_property (G_OBJECT_CLASS (klass), PROP_BITRATE,
       g_param_spec_int ("bitrate", "Encoding Bit-rate",
           "Specify an encoding bit-rate (in bps). (0 = automatic)",
           0, G_MAXINT, DEFAULT_BITRATE,
-          G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
+          G_PARAM_CONSTRUCT | G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
   g_object_class_install_property (gobject_class, PROP_MODE,
       g_param_spec_enum ("mode", "Mode", "The encoding mode",
           GST_TYPE_SPEEX_ENC_MODE, GST_SPEEX_ENC_MODE_AUTO,
-          G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
+          G_PARAM_CONSTRUCT | G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
   g_object_class_install_property (G_OBJECT_CLASS (klass), PROP_VBR,
       g_param_spec_boolean ("vbr", "VBR",
           "Enable variable bit-rate", DEFAULT_VBR,
-          G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
+          G_PARAM_CONSTRUCT | G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
   g_object_class_install_property (G_OBJECT_CLASS (klass), PROP_ABR,
       g_param_spec_int ("abr", "ABR",
           "Enable average bit-rate (0 = disabled)",
           0, G_MAXINT, DEFAULT_ABR,
-          G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
+          G_PARAM_CONSTRUCT | G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
   g_object_class_install_property (G_OBJECT_CLASS (klass), PROP_VAD,
       g_param_spec_boolean ("vad", "VAD",
           "Enable voice activity detection", DEFAULT_VAD,
-          G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
+          G_PARAM_CONSTRUCT | G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
   g_object_class_install_property (G_OBJECT_CLASS (klass), PROP_DTX,
       g_param_spec_boolean ("dtx", "DTX",
           "Enable discontinuous transmission", DEFAULT_DTX,
-          G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
+          G_PARAM_CONSTRUCT | G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
   g_object_class_install_property (G_OBJECT_CLASS (klass), PROP_COMPLEXITY,
       g_param_spec_int ("complexity", "Complexity",
           "Set encoding complexity",
           0, G_MAXINT, DEFAULT_COMPLEXITY,
-          G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
+          G_PARAM_CONSTRUCT | G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
   g_object_class_install_property (G_OBJECT_CLASS (klass), PROP_NFRAMES,
       g_param_spec_int ("nframes", "NFrames",
           "Number of frames per buffer",
           0, G_MAXINT, DEFAULT_NFRAMES,
-          G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
+          G_PARAM_CONSTRUCT | G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
   g_object_class_install_property (G_OBJECT_CLASS (klass), PROP_LAST_MESSAGE,
       g_param_spec_string ("last-message", "last-message",
           "The last status message", NULL,
           G_PARAM_READABLE | G_PARAM_STATIC_STRINGS));
 
-  gstelement_class->change_state =
-      GST_DEBUG_FUNCPTR (gst_speex_enc_change_state);
-
   gst_element_class_add_pad_template (gstelement_class,
       gst_static_pad_template_get (&src_factory));
   gst_element_class_add_pad_template (gstelement_class,
@@ -233,166 +223,49 @@ gst_speex_enc_finalize (GObject * object)
   enc = GST_SPEEX_ENC (object);
 
   g_free (enc->last_message);
-  g_object_unref (enc->adapter);
 
   G_OBJECT_CLASS (parent_class)->finalize (object);
 }
 
-static gboolean
-gst_speex_enc_sink_setcaps (GstPad * pad, GstCaps * caps)
-{
-  GstSpeexEnc *enc;
-  GstStructure *structure;
-
-  enc = GST_SPEEX_ENC (GST_PAD_PARENT (pad));
-  enc->setup = FALSE;
-
-  structure = gst_caps_get_structure (caps, 0);
-  gst_structure_get_int (structure, "channels", &enc->channels);
-  gst_structure_get_int (structure, "rate", &enc->rate);
-
-  gst_speex_enc_setup (enc);
-
-  return enc->setup;
-}
-
-
-static GstCaps *
-gst_speex_enc_sink_getcaps (GstPad * pad, GstCaps * filter)
+static void
+gst_speex_enc_init (GstSpeexEnc * enc)
 {
-  GstCaps *caps = gst_caps_copy (gst_pad_get_pad_template_caps (pad));
-  GstCaps *peercaps = NULL;
-  GstSpeexEnc *enc = GST_SPEEX_ENC (gst_pad_get_parent_element (pad));
-
-  peercaps = gst_pad_peer_get_caps (enc->srcpad, filter);
-
-  if (peercaps) {
-    if (!gst_caps_is_empty (peercaps) && !gst_caps_is_any (peercaps)) {
-      GstStructure *ps = gst_caps_get_structure (peercaps, 0);
-      GstStructure *s = gst_caps_get_structure (caps, 0);
-      gint rate, channels;
-
-      if (gst_structure_get_int (ps, "rate", &rate)) {
-        gst_structure_fixate_field_nearest_int (s, "rate", rate);
-      }
-
-      if (gst_structure_get_int (ps, "channels", &channels)) {
-        gst_structure_fixate_field_nearest_int (s, "channels", channels);
-      }
-    }
-    gst_caps_unref (peercaps);
-  }
-
-  gst_object_unref (enc);
+  GstAudioEncoder *benc = GST_AUDIO_ENCODER (enc);
 
-  return caps;
+  /* arrange granulepos marking (and required perfect ts) */
+  gst_audio_encoder_set_mark_granule (benc, TRUE);
+  gst_audio_encoder_set_perfect_timestamp (benc, TRUE);
 }
 
-
 static gboolean
-gst_speex_enc_convert_src (GstPad * pad, GstFormat src_format, gint64 src_value,
-    GstFormat dest_format, gint64 * dest_value)
+gst_speex_enc_start (GstAudioEncoder * benc)
 {
-  gboolean res = TRUE;
-  GstSpeexEnc *enc;
-  gint64 avg;
-
-  enc = GST_SPEEX_ENC (GST_PAD_PARENT (pad));
+  GstSpeexEnc *enc = GST_SPEEX_ENC (benc);
 
-  if (enc->samples_in == 0 || enc->bytes_out == 0 || enc->rate == 0)
-    return FALSE;
-
-  avg = (enc->bytes_out * enc->rate) / (enc->samples_in);
+  GST_DEBUG_OBJECT (enc, "start");
+  speex_bits_init (&enc->bits);
+  enc->tags = gst_tag_list_new ();
+  enc->header_sent = FALSE;
 
-  switch (src_format) {
-    case GST_FORMAT_BYTES:
-      switch (dest_format) {
-        case GST_FORMAT_TIME:
-          *dest_value = src_value * GST_SECOND / avg;
-          break;
-        default:
-          res = FALSE;
-      }
-      break;
-    case GST_FORMAT_TIME:
-      switch (dest_format) {
-        case GST_FORMAT_BYTES:
-          *dest_value = src_value * avg / GST_SECOND;
-          break;
-        default:
-          res = FALSE;
-      }
-      break;
-    default:
-      res = FALSE;
-  }
-  return res;
+  return TRUE;
 }
 
 static gboolean
-gst_speex_enc_convert_sink (GstPad * pad, GstFormat src_format,
-    gint64 src_value, GstFormat * dest_format, gint64 * dest_value)
+gst_speex_enc_stop (GstAudioEncoder * benc)
 {
-  gboolean res = TRUE;
-  guint scale = 1;
-  gint bytes_per_sample;
-  GstSpeexEnc *enc;
+  GstSpeexEnc *enc = GST_SPEEX_ENC (benc);
 
-  enc = GST_SPEEX_ENC (GST_PAD_PARENT (pad));
-
-  bytes_per_sample = enc->channels * 2;
-
-  switch (src_format) {
-    case GST_FORMAT_BYTES:
-      switch (*dest_format) {
-        case GST_FORMAT_DEFAULT:
-          if (bytes_per_sample == 0)
-            return FALSE;
-          *dest_value = src_value / bytes_per_sample;
-          break;
-        case GST_FORMAT_TIME:
-        {
-          gint byterate = bytes_per_sample * enc->rate;
-
-          if (byterate == 0)
-            return FALSE;
-          *dest_value = src_value * GST_SECOND / byterate;
-          break;
-        }
-        default:
-          res = FALSE;
-      }
-      break;
-    case GST_FORMAT_DEFAULT:
-      switch (*dest_format) {
-        case GST_FORMAT_BYTES:
-          *dest_value = src_value * bytes_per_sample;
-          break;
-        case GST_FORMAT_TIME:
-          if (enc->rate == 0)
-            return FALSE;
-          *dest_value = src_value * GST_SECOND / enc->rate;
-          break;
-        default:
-          res = FALSE;
-      }
-      break;
-    case GST_FORMAT_TIME:
-      switch (*dest_format) {
-        case GST_FORMAT_BYTES:
-          scale = bytes_per_sample;
-          /* fallthrough */
-        case GST_FORMAT_DEFAULT:
-          *dest_value = src_value * scale * enc->rate / GST_SECOND;
-          break;
-        default:
-          res = FALSE;
-      }
-      break;
-    default:
-      res = FALSE;
+  GST_DEBUG_OBJECT (enc, "stop");
+  enc->header_sent = FALSE;
+  if (enc->state) {
+    speex_encoder_destroy (enc->state);
+    enc->state = NULL;
   }
-  return res;
+  speex_bits_destroy (&enc->bits);
+  gst_tag_list_free (enc->tags);
+  enc->tags = NULL;
+
+  return TRUE;
 }
 
 static gint64
@@ -405,185 +278,45 @@ gst_speex_enc_get_latency (GstSpeexEnc * enc)
     return 34 * GST_MSECOND;
 }
 
-static const GstQueryType *
-gst_speex_enc_get_query_types (GstPad * pad)
-{
-  static const GstQueryType gst_speex_enc_src_query_types[] = {
-    GST_QUERY_POSITION,
-    GST_QUERY_DURATION,
-    GST_QUERY_CONVERT,
-    GST_QUERY_LATENCY,
-    0
-  };
-
-  return gst_speex_enc_src_query_types;
-}
-
 static gboolean
-gst_speex_enc_src_query (GstPad * pad, GstQuery * query)
+gst_speex_enc_set_format (GstAudioEncoder * benc, GstAudioInfo * info)
 {
-  gboolean res = TRUE;
   GstSpeexEnc *enc;
 
-  enc = GST_SPEEX_ENC (gst_pad_get_parent (pad));
-
-  switch (GST_QUERY_TYPE (query)) {
-    case GST_QUERY_POSITION:
-    {
-      GstFormat req_fmt;
-      gint64 pos, val;
-
-      gst_query_parse_position (query, &req_fmt, NULL);
-      if ((res = gst_pad_query_peer_position (enc->sinkpad, req_fmt, &val))) {
-        gst_query_set_position (query, req_fmt, val);
-        break;
-      }
-
-      res = gst_pad_query_peer_position (enc->sinkpad, GST_FORMAT_TIME, &pos);
-      if (!res)
-        break;
-
-      if ((res =
-              gst_pad_query_peer_convert (enc->sinkpad, GST_FORMAT_TIME, pos,
-                  req_fmt, &val))) {
-        gst_query_set_position (query, req_fmt, val);
-      }
-      break;
-    }
-    case GST_QUERY_DURATION:
-    {
-      GstFormat req_fmt;
-      gint64 dur, val;
-
-      gst_query_parse_duration (query, &req_fmt, NULL);
-      if ((res = gst_pad_query_peer_duration (enc->sinkpad, req_fmt, &val))) {
-        gst_query_set_duration (query, req_fmt, val);
-        break;
-      }
+  enc = GST_SPEEX_ENC (benc);
 
-      res = gst_pad_query_peer_duration (enc->sinkpad, GST_FORMAT_TIME, &dur);
-      if (!res)
-        break;
+  enc->channels = GST_AUDIO_INFO_CHANNELS (info);
+  enc->rate = GST_AUDIO_INFO_RATE (info);
 
-      if ((res =
-              gst_pad_query_peer_convert (enc->sinkpad, GST_FORMAT_TIME, dur,
-                  req_fmt, &val))) {
-        gst_query_set_duration (query, req_fmt, val);
-      }
-      break;
-    }
-    case GST_QUERY_CONVERT:
-    {
-      GstFormat src_fmt, dest_fmt;
-      gint64 src_val, dest_val;
-
-      gst_query_parse_convert (query, &src_fmt, &src_val, &dest_fmt, &dest_val);
-      if (!(res = gst_speex_enc_convert_src (pad, src_fmt, src_val, dest_fmt,
-                  &dest_val)))
-        goto error;
-      gst_query_set_convert (query, src_fmt, src_val, dest_fmt, dest_val);
-      break;
-    }
-    case GST_QUERY_LATENCY:
-    {
-      gboolean live;
-      GstClockTime min_latency, max_latency;
-      gint64 latency;
-
-      if ((res = gst_pad_peer_query (enc->sinkpad, query))) {
-        gst_query_parse_latency (query, &live, &min_latency, &max_latency);
-        GST_LOG_OBJECT (pad, "Upstream latency: %" GST_PTR_FORMAT, query);
-
-        latency = gst_speex_enc_get_latency (enc);
-
-        /* add our latency */
-        min_latency += latency;
-        if (max_latency != -1)
-          max_latency += latency;
-
-        gst_query_set_latency (query, live, min_latency, max_latency);
-        GST_LOG_OBJECT (pad, "Adjusted latency: %" GST_PTR_FORMAT, query);
-      }
-      break;
-    }
-    default:
-      res = gst_pad_peer_query (enc->sinkpad, query);
-      break;
+  /* handle reconfigure */
+  if (enc->state) {
+    speex_encoder_destroy (enc->state);
+    enc->state = NULL;
   }
 
-error:
-
-  gst_object_unref (enc);
-
-  return res;
-}
+  if (!gst_speex_enc_setup (enc))
+    return FALSE;
 
-static gboolean
-gst_speex_enc_sink_query (GstPad * pad, GstQuery * query)
-{
-  gboolean res = TRUE;
+  /* feedback to base class */
+  gst_audio_encoder_set_latency (benc,
+      gst_speex_enc_get_latency (enc), gst_speex_enc_get_latency (enc));
+  gst_audio_encoder_set_lookahead (benc, enc->lookahead);
 
-  switch (GST_QUERY_TYPE (query)) {
-    case GST_QUERY_CONVERT:
-    {
-      GstFormat src_fmt, dest_fmt;
-      gint64 src_val, dest_val;
-
-      gst_query_parse_convert (query, &src_fmt, &src_val, &dest_fmt, &dest_val);
-      if (!(res =
-              gst_speex_enc_convert_sink (pad, src_fmt, src_val, &dest_fmt,
-                  &dest_val)))
-        goto error;
-      gst_query_set_convert (query, src_fmt, src_val, dest_fmt, dest_val);
-      break;
-    }
-    default:
-      res = gst_pad_query_default (pad, query);
-      break;
+  if (enc->nframes == 0) {
+    /* as many frames as available input allows */
+    gst_audio_encoder_set_frame_samples_min (benc, enc->frame_size);
+    gst_audio_encoder_set_frame_samples_max (benc, enc->frame_size);
+    gst_audio_encoder_set_frame_max (benc, 0);
+  } else {
+    /* exactly as many frames as configured */
+    gst_audio_encoder_set_frame_samples_min (benc,
+        enc->frame_size * enc->nframes);
+    gst_audio_encoder_set_frame_samples_max (benc,
+        enc->frame_size * enc->nframes);
+    gst_audio_encoder_set_frame_max (benc, 1);
   }
 
-error:
-  return res;
-}
-
-static void
-gst_speex_enc_init (GstSpeexEnc * enc)
-{
-  enc->sinkpad = gst_pad_new_from_static_template (&sink_factory, "sink");
-  gst_element_add_pad (GST_ELEMENT (enc), enc->sinkpad);
-  gst_pad_set_event_function (enc->sinkpad,
-      GST_DEBUG_FUNCPTR (gst_speex_enc_sink_event));
-  gst_pad_set_chain_function (enc->sinkpad,
-      GST_DEBUG_FUNCPTR (gst_speex_enc_chain));
-  gst_pad_set_getcaps_function (enc->sinkpad,
-      GST_DEBUG_FUNCPTR (gst_speex_enc_sink_getcaps));
-  gst_pad_set_query_function (enc->sinkpad,
-      GST_DEBUG_FUNCPTR (gst_speex_enc_sink_query));
-
-  enc->srcpad = gst_pad_new_from_static_template (&src_factory, "src");
-  gst_pad_set_query_function (enc->srcpad,
-      GST_DEBUG_FUNCPTR (gst_speex_enc_src_query));
-  gst_pad_set_query_type_function (enc->srcpad,
-      GST_DEBUG_FUNCPTR (gst_speex_enc_get_query_types));
-  gst_element_add_pad (GST_ELEMENT (enc), enc->srcpad);
-
-  enc->channels = -1;
-  enc->rate = -1;
-
-  enc->quality = DEFAULT_QUALITY;
-  enc->bitrate = DEFAULT_BITRATE;
-  enc->mode = DEFAULT_MODE;
-  enc->vbr = DEFAULT_VBR;
-  enc->abr = DEFAULT_ABR;
-  enc->vad = DEFAULT_VAD;
-  enc->dtx = DEFAULT_DTX;
-  enc->complexity = DEFAULT_COMPLEXITY;
-  enc->nframes = DEFAULT_NFRAMES;
-
-  enc->setup = FALSE;
-  enc->header_sent = FALSE;
-
-  enc->adapter = gst_adapter_new ();
+  return TRUE;
 }
 
 static GstBuffer *
@@ -610,7 +343,7 @@ gst_speex_enc_create_metadata_buffer (GstSpeexEnc * enc)
       0, "Encoded with GStreamer Speexenc");
   gst_tag_list_free (merged_tags);
 
-  GST_BUFFER_OFFSET (comments) = enc->bytes_out;
+  GST_BUFFER_OFFSET (comments) = 0;
   GST_BUFFER_OFFSET_END (comments) = 0;
 
   return comments;
@@ -628,8 +361,6 @@ gst_speex_enc_set_last_msg (GstSpeexEnc * enc, const gchar * msg)
 static gboolean
 gst_speex_enc_setup (GstSpeexEnc * enc)
 {
-  enc->setup = FALSE;
-
   switch (enc->mode) {
     case GST_SPEEX_ENC_MODE_UWB:
       GST_LOG_OBJECT (enc, "configuring for requested UWB mode");
@@ -746,105 +477,27 @@ gst_speex_enc_setup (GstSpeexEnc * enc)
   GST_LOG_OBJECT (enc, "we have frame size %d, lookahead %d", enc->frame_size,
       enc->lookahead);
 
-  enc->setup = TRUE;
-
   return TRUE;
 }
 
-/* prepare a buffer for transmission */
-static GstBuffer *
-gst_speex_enc_buffer_from_data (GstSpeexEnc * enc, guchar * data,
-    gint data_len, guint64 granulepos)
-{
-  GstBuffer *outbuf;
-
-  outbuf = gst_buffer_new_and_alloc (data_len);
-  gst_buffer_fill (outbuf, 0, data, data_len);
-  GST_BUFFER_OFFSET (outbuf) = enc->bytes_out;
-  GST_BUFFER_OFFSET_END (outbuf) = granulepos;
-
-  GST_LOG_OBJECT (enc, "encoded buffer of %d bytes", data_len);
-  return outbuf;
-}
-
-
-/* push out the buffer and do internal bookkeeping */
+/* push out the buffer */
 static GstFlowReturn
 gst_speex_enc_push_buffer (GstSpeexEnc * enc, GstBuffer * buffer)
 {
-  guint size;
-
-  size = gst_buffer_get_size (buffer);
-  enc->bytes_out += size;
-
-  GST_DEBUG_OBJECT (enc, "pushing output buffer of size %u", size);
-
-  return gst_pad_push (enc->srcpad, buffer);
-}
-
-static GstCaps *
-gst_speex_enc_set_header_on_caps (GstCaps * caps, GstBuffer * buf1,
-    GstBuffer * buf2)
-{
-  GstStructure *structure = NULL;
-  GstBuffer *buf;
-  GValue array = { 0 };
-  GValue value = { 0 };
-
-  caps = gst_caps_make_writable (caps);
-  structure = gst_caps_get_structure (caps, 0);
+  GST_DEBUG_OBJECT (enc, "pushing output buffer of size %u",
+      gst_buffer_get_size (buffer));
 
-  g_assert (gst_buffer_is_writable (buf1));
-  g_assert (gst_buffer_is_writable (buf2));
-
-  /* mark buffers */
-  GST_BUFFER_FLAG_SET (buf1, GST_BUFFER_FLAG_IN_CAPS);
-  GST_BUFFER_FLAG_SET (buf2, GST_BUFFER_FLAG_IN_CAPS);
-
-  /* put buffers in a fixed list */
-  g_value_init (&array, GST_TYPE_ARRAY);
-  g_value_init (&value, GST_TYPE_BUFFER);
-  buf = gst_buffer_copy (buf1);
-  gst_value_set_buffer (&value, buf);
-  gst_buffer_unref (buf);
-  gst_value_array_append_value (&array, &value);
-  g_value_unset (&value);
-  g_value_init (&value, GST_TYPE_BUFFER);
-  buf = gst_buffer_copy (buf2);
-  gst_value_set_buffer (&value, buf);
-  gst_buffer_unref (buf);
-  gst_value_array_append_value (&array, &value);
-  gst_structure_set_value (structure, "streamheader", &array);
-  g_value_unset (&value);
-  g_value_unset (&array);
-
-  return caps;
+  return gst_pad_push (GST_AUDIO_ENCODER_SRC_PAD (enc), buffer);
 }
 
-
 static gboolean
-gst_speex_enc_sink_event (GstPad * pad, GstEvent * event)
+gst_speex_enc_sink_event (GstAudioEncoder * benc, GstEvent * event)
 {
-  gboolean res = TRUE;
   GstSpeexEnc *enc;
 
-  enc = GST_SPEEX_ENC (gst_pad_get_parent (pad));
+  enc = GST_SPEEX_ENC (benc);
 
   switch (GST_EVENT_TYPE (event)) {
-    case GST_EVENT_CAPS:
-    {
-      GstCaps *caps;
-
-      gst_event_parse_caps (event, &caps);
-      res = gst_speex_enc_sink_setcaps (pad, caps);
-      gst_event_unref (event);
-      break;
-    }
-    case GST_EVENT_EOS:
-      if (enc->setup)
-        gst_speex_enc_encode (enc, TRUE);
-      res = gst_pad_event_default (pad, event);
-      break;
     case GST_EVENT_TAG:
     {
       if (enc->tags) {
@@ -856,116 +509,165 @@ gst_speex_enc_sink_event (GstPad * pad, GstEvent * event)
       } else {
         g_assert_not_reached ();
       }
-      res = gst_pad_event_default (pad, event);
       break;
     }
     default:
-      res = gst_pad_event_default (pad, event);
       break;
   }
 
-  gst_object_unref (enc);
-
-  return res;
+  /* we only peeked, let base class handle it */
+  return FALSE;
 }
 
 static GstFlowReturn
-gst_speex_enc_encode (GstSpeexEnc * enc, gboolean flush)
+gst_speex_enc_encode (GstSpeexEnc * enc, GstBuffer * buf)
 {
   gint frame_size = enc->frame_size;
-  gint bytes = frame_size * 2 * enc->channels;
+  gint bytes = frame_size * 2 * enc->channels, samples;
+  gint outsize, written, dtx_ret = 0;
+  guint8 *data, *bdata, *outdata;
+  gsize bsize, size;
+  GstBuffer *outbuf;
   GstFlowReturn ret = GST_FLOW_OK;
 
-  if (flush && gst_adapter_available (enc->adapter) % bytes != 0) {
-    guint diff = gst_adapter_available (enc->adapter) % bytes;
-    GstBuffer *buf = gst_buffer_new_and_alloc (diff);
-    gst_buffer_memset (buf, 0, 0, diff);
-    gst_adapter_push (enc->adapter, buf);
-  }
+  if (G_LIKELY (buf)) {
+    bdata = gst_buffer_map (buf, &bsize, NULL, GST_MAP_READ);
+
+    if (G_UNLIKELY (bsize % bytes)) {
+      GST_DEBUG_OBJECT (enc, "draining; adding silence samples");
 
-  while (gst_adapter_available (enc->adapter) >= bytes) {
-    gint16 *data;
-    gint outsize, written, dtx_ret;
-    GstBuffer *outbuf;
-    gchar *outdata;
+      size = ((bsize / bytes) + 1) * bytes;
+      data = g_malloc0 (size);
+      memcpy (data, bdata, bsize);
+      gst_buffer_unmap (buf, bdata, bsize);
+      bdata = NULL;
+    } else {
+      data = bdata;
+      size = bsize;
+    }
+  } else {
+    GST_DEBUG_OBJECT (enc, "nothing to drain");
+    goto done;
+  }
 
-    data = (gint16 *) gst_adapter_take (enc->adapter, bytes);
+  samples = size / (2 * enc->channels);
+  speex_bits_reset (&enc->bits);
 
-    enc->samples_in += frame_size;
+  /* FIXME what about dropped samples if DTS enabled ?? */
 
+  while (size) {
     GST_DEBUG_OBJECT (enc, "encoding %d samples (%d bytes)", frame_size, bytes);
 
     if (enc->channels == 2) {
-      speex_encode_stereo_int (data, frame_size, &enc->bits);
+      speex_encode_stereo_int ((gint16 *) data, frame_size, &enc->bits);
     }
-    dtx_ret = speex_encode_int (enc->state, data, &enc->bits);
-
-    g_free (data);
+    dtx_ret += speex_encode_int (enc->state, (gint16 *) data, &enc->bits);
 
-    enc->frameno++;
-    enc->frameno_out++;
+    data += bytes;
+    size -= bytes;
+  }
 
-    if ((enc->frameno % enc->nframes) != 0)
-      continue;
+  speex_bits_insert_terminator (&enc->bits);
+  outsize = speex_bits_nbytes (&enc->bits);
 
-    speex_bits_insert_terminator (&enc->bits);
-    outsize = speex_bits_nbytes (&enc->bits);
+  if (bdata)
+    gst_buffer_unmap (buf, bdata, bsize);
 
 #if 0
-    ret = gst_pad_alloc_buffer_and_set_caps (enc->srcpad,
-        GST_BUFFER_OFFSET_NONE, outsize, GST_PAD_CAPS (enc->srcpad), &outbuf);
-    if ((GST_FLOW_OK != ret))
-      goto done;
+  ret = gst_pad_alloc_buffer_and_set_caps (GST_AUDIO_ENCODER_SRC_PAD (enc),
+      GST_BUFFER_OFFSET_NONE, outsize,
+      GST_PAD_CAPS (GST_AUDIO_ENCODER_SRC_PAD (enc)), &outbuf);
+
+  if ((GST_FLOW_OK != ret))
+    goto done;
 #endif
-    outbuf = gst_buffer_new_allocate (NULL, outsize, 0);
+  outbuf = gst_buffer_new_allocate (NULL, outsize, 0);
+  outdata = gst_buffer_map (outbuf, NULL, NULL, GST_MAP_WRITE);
 
-    outdata = gst_buffer_map (outbuf, NULL, NULL, GST_MAP_WRITE);
-    written = speex_bits_write (&enc->bits, outdata, outsize);
+  written = speex_bits_write (&enc->bits, (gchar *) outdata, outsize);
 
-    if (G_UNLIKELY (written != outsize)) {
-      GST_ERROR_OBJECT (enc, "short write: %d < %d bytes", written, outsize);
-    }
-    gst_buffer_unmap (outbuf, outdata, written);
+  if (G_UNLIKELY (written < outsize)) {
+    GST_ERROR_OBJECT (enc, "short write: %d < %d bytes", written, outsize);
+  } else if (G_UNLIKELY (written > outsize)) {
+    GST_ERROR_OBJECT (enc, "overrun: %d > %d bytes", written, outsize);
+    written = outsize;
+  }
+  gst_buffer_unmap (outbuf, outdata, written);
+
+  if (!dtx_ret)
+    GST_BUFFER_FLAG_SET (outbuf, GST_BUFFER_FLAG_GAP);
 
-    speex_bits_reset (&enc->bits);
+  ret = gst_audio_encoder_finish_frame (GST_AUDIO_ENCODER (enc),
+      outbuf, samples);
 
-    if (!dtx_ret)
-      GST_BUFFER_FLAG_SET (outbuf, GST_BUFFER_FLAG_GAP);
+done:
+  return ret;
+}
 
-    GST_BUFFER_TIMESTAMP (outbuf) = enc->start_ts +
-        gst_util_uint64_scale_int ((enc->frameno_out -
-            enc->nframes) * frame_size - enc->lookahead, GST_SECOND, enc->rate);
-    GST_BUFFER_DURATION (outbuf) =
-        gst_util_uint64_scale_int (frame_size * enc->nframes, GST_SECOND,
-        enc->rate);
-    /* set gp time and granulepos; see gst-plugins-base/ext/ogg/README */
-    GST_BUFFER_OFFSET_END (outbuf) = enc->granulepos_offset +
-        ((enc->frameno_out) * frame_size - enc->lookahead);
-    GST_BUFFER_OFFSET (outbuf) =
-        gst_util_uint64_scale_int (GST_BUFFER_OFFSET_END (outbuf), GST_SECOND,
-        enc->rate);
+/*
+ * (really really) FIXME: move into core (dixit tpm)
+ */
+/**
+ * _gst_caps_set_buffer_array:
+ * @caps: a #GstCaps
+ * @field: field in caps to set
+ * @buf: header buffers
+ *
+ * Adds given buffers to an array of buffers set as the given @field
+ * on the given @caps.  List of buffer arguments must be NULL-terminated.
+ *
+ * Returns: input caps with a streamheader field added, or NULL if some error
+ */
+static GstCaps *
+_gst_caps_set_buffer_array (GstCaps * caps, const gchar * field,
+    GstBuffer * buf, ...)
+{
+  GstStructure *structure = NULL;
+  va_list va;
+  GValue array = { 0 };
+  GValue value = { 0 };
 
-    ret = gst_speex_enc_push_buffer (enc, outbuf);
+  g_return_val_if_fail (caps != NULL, NULL);
+  g_return_val_if_fail (gst_caps_is_fixed (caps), NULL);
+  g_return_val_if_fail (field != NULL, NULL);
 
-    if ((GST_FLOW_OK != ret) && (GST_FLOW_NOT_LINKED != ret))
-      goto done;
+  caps = gst_caps_make_writable (caps);
+  structure = gst_caps_get_structure (caps, 0);
+
+  g_value_init (&array, GST_TYPE_ARRAY);
+
+  va_start (va, buf);
+  /* put buffers in a fixed list */
+  while (buf) {
+    g_assert (gst_buffer_is_writable (buf));
+
+    /* mark buffer */
+    GST_BUFFER_FLAG_SET (buf, GST_BUFFER_FLAG_IN_CAPS);
+
+    g_value_init (&value, GST_TYPE_BUFFER);
+    buf = gst_buffer_copy (buf);
+    GST_BUFFER_FLAG_SET (buf, GST_BUFFER_FLAG_IN_CAPS);
+    gst_value_set_buffer (&value, buf);
+    gst_buffer_unref (buf);
+    gst_value_array_append_value (&array, &value);
+    g_value_unset (&value);
+
+    buf = va_arg (va, GstBuffer *);
   }
 
-done:
+  gst_structure_set_value (structure, field, &array);
+  g_value_unset (&array);
 
-  return ret;
+  return caps;
 }
 
 static GstFlowReturn
-gst_speex_enc_chain (GstPad * pad, GstBuffer * buf)
+gst_speex_enc_handle_frame (GstAudioEncoder * benc, GstBuffer * buf)
 {
   GstSpeexEnc *enc;
   GstFlowReturn ret = GST_FLOW_OK;
 
-  enc = GST_SPEEX_ENC (GST_PAD_PARENT (pad));
-
-  if (!enc->setup)
-    goto not_setup;
+  enc = GST_SPEEX_ENC (benc);
 
   if (!enc->header_sent) {
     /* Speex streams begin with two headers; the initial header (with
@@ -981,24 +683,20 @@ gst_speex_enc_chain (GstPad * pad, GstBuffer * buf)
 
     /* create header buffer */
     data = (guint8 *) speex_header_to_packet (&enc->header, &data_len);
-    buf1 = gst_speex_enc_buffer_from_data (enc, data, data_len, 0);
-    free (data);
+    buf1 = gst_buffer_new_wrapped (data, data_len);
+    GST_BUFFER_OFFSET_END (buf1) = 0;
+    GST_BUFFER_OFFSET (buf1) = 0;
 
     /* create comment buffer */
     buf2 = gst_speex_enc_create_metadata_buffer (enc);
 
     /* mark and put on caps */
-    caps = gst_pad_get_caps (enc->srcpad, NULL);
-    caps = gst_speex_enc_set_header_on_caps (caps, buf1, buf2);
-
-    gst_caps_set_simple (caps,
-        "rate", G_TYPE_INT, enc->rate,
+    caps = gst_caps_new_simple ("audio/x-speex", "rate", G_TYPE_INT, enc->rate,
         "channels", G_TYPE_INT, enc->channels, NULL);
+    caps = _gst_caps_set_buffer_array (caps, "streamheader", buf1, buf2, NULL);
 
     /* negotiate with these caps */
     GST_DEBUG_OBJECT (enc, "here are the caps: %" GST_PTR_FORMAT, caps);
-    gst_pad_set_caps (enc->srcpad, caps);
-    gst_caps_unref (caps);
 
     /* push out buffers */
     ret = gst_speex_enc_push_buffer (enc, buf1);
@@ -1013,102 +711,16 @@ gst_speex_enc_chain (GstPad * pad, GstBuffer * buf)
     if (ret != GST_FLOW_OK)
       goto done;
 
-    speex_bits_reset (&enc->bits);
-
     enc->header_sent = TRUE;
   }
 
-  /* Save the timestamp of the first buffer. This will be later
-   * used as offset for all following buffers */
-  if (enc->start_ts == GST_CLOCK_TIME_NONE) {
-    if (GST_BUFFER_TIMESTAMP_IS_VALID (buf)) {
-      enc->start_ts = GST_BUFFER_TIMESTAMP (buf);
-      enc->granulepos_offset = gst_util_uint64_scale
-          (GST_BUFFER_TIMESTAMP (buf), enc->rate, GST_SECOND);
-    } else {
-      enc->start_ts = 0;
-      enc->granulepos_offset = 0;
-    }
-  }
-
-  /* Check if we have a continous stream, if not drop some samples or the buffer or
-   * insert some silence samples */
-  if (enc->next_ts != GST_CLOCK_TIME_NONE &&
-      GST_BUFFER_TIMESTAMP_IS_VALID (buf) &&
-      GST_BUFFER_TIMESTAMP (buf) < enc->next_ts) {
-    guint64 diff = enc->next_ts - GST_BUFFER_TIMESTAMP (buf);
-    guint64 diff_bytes;
-
-    GST_WARNING_OBJECT (enc, "Buffer is older than previous "
-        "timestamp + duration (%" GST_TIME_FORMAT "< %" GST_TIME_FORMAT
-        "), cannot handle. Clipping buffer.",
-        GST_TIME_ARGS (GST_BUFFER_TIMESTAMP (buf)),
-        GST_TIME_ARGS (enc->next_ts));
-
-    diff_bytes = GST_CLOCK_TIME_TO_FRAMES (diff, enc->rate) * enc->channels * 2;
-    if (diff_bytes >= gst_buffer_get_size (buf)) {
-      gst_buffer_unref (buf);
-      return GST_FLOW_OK;
-    }
-    buf = gst_buffer_make_writable (buf);
-    gst_buffer_resize (buf, diff_bytes, -1);
-
-    GST_BUFFER_TIMESTAMP (buf) += diff;
-    if (GST_BUFFER_DURATION_IS_VALID (buf))
-      GST_BUFFER_DURATION (buf) -= diff;
-  }
+  GST_DEBUG_OBJECT (enc, "received buffer %p of %u bytes", buf,
+      buf ? gst_buffer_get_size (buf) : 0);
 
-  if (enc->next_ts != GST_CLOCK_TIME_NONE
-      && GST_BUFFER_TIMESTAMP_IS_VALID (buf)) {
-    guint64 max_diff =
-        gst_util_uint64_scale (enc->frame_size, GST_SECOND, enc->rate);
-
-    if (GST_BUFFER_TIMESTAMP (buf) != enc->next_ts &&
-        GST_BUFFER_TIMESTAMP (buf) - enc->next_ts > max_diff) {
-      GST_WARNING_OBJECT (enc,
-          "Discontinuity detected: %" G_GUINT64_FORMAT " > %" G_GUINT64_FORMAT,
-          GST_BUFFER_TIMESTAMP (buf) - enc->next_ts, max_diff);
-
-      gst_speex_enc_encode (enc, TRUE);
-
-      enc->frameno_out = 0;
-      enc->start_ts = GST_BUFFER_TIMESTAMP (buf);
-      enc->granulepos_offset = gst_util_uint64_scale
-          (GST_BUFFER_TIMESTAMP (buf), enc->rate, GST_SECOND);
-    }
-  }
-
-  if (GST_BUFFER_TIMESTAMP_IS_VALID (buf)
-      && GST_BUFFER_DURATION_IS_VALID (buf))
-    enc->next_ts = GST_BUFFER_TIMESTAMP (buf) + GST_BUFFER_DURATION (buf);
-  else
-    enc->next_ts = GST_CLOCK_TIME_NONE;
-
-  GST_DEBUG_OBJECT (enc, "received buffer of %u bytes",
-      gst_buffer_get_size (buf));
-
-  /* push buffer to adapter */
-  gst_adapter_push (enc->adapter, buf);
-  buf = NULL;
-
-  ret = gst_speex_enc_encode (enc, FALSE);
+  ret = gst_speex_enc_encode (enc, buf);
 
 done:
-
-  if (buf)
-    gst_buffer_unref (buf);
-
   return ret;
-
-  /* ERRORS */
-not_setup:
-  {
-    GST_ELEMENT_ERROR (enc, CORE, NEGOTIATION, (NULL),
-        ("encoder not initialized (input is not audio?)"));
-    ret = GST_FLOW_NOT_NEGOTIATED;
-    goto done;
-  }
-
 }
 
 
@@ -1198,54 +810,3 @@ gst_speex_enc_set_property (GObject * object, guint prop_id,
       break;
   }
 }
-
-static GstStateChangeReturn
-gst_speex_enc_change_state (GstElement * element, GstStateChange transition)
-{
-  GstSpeexEnc *enc = GST_SPEEX_ENC (element);
-  GstStateChangeReturn res;
-
-  switch (transition) {
-    case GST_STATE_CHANGE_NULL_TO_READY:
-      enc->tags = gst_tag_list_new ();
-      break;
-    case GST_STATE_CHANGE_READY_TO_PAUSED:
-      speex_bits_init (&enc->bits);
-      enc->frameno = 0;
-      enc->frameno_out = 0;
-      enc->samples_in = 0;
-      enc->start_ts = GST_CLOCK_TIME_NONE;
-      enc->next_ts = GST_CLOCK_TIME_NONE;
-      enc->granulepos_offset = 0;
-      break;
-    case GST_STATE_CHANGE_PAUSED_TO_PLAYING:
-      /* fall through */
-    default:
-      break;
-  }
-
-  res = GST_ELEMENT_CLASS (parent_class)->change_state (element, transition);
-  if (res == GST_STATE_CHANGE_FAILURE)
-    return res;
-
-  switch (transition) {
-    case GST_STATE_CHANGE_PLAYING_TO_PAUSED:
-      break;
-    case GST_STATE_CHANGE_PAUSED_TO_READY:
-      enc->setup = FALSE;
-      enc->header_sent = FALSE;
-      if (enc->state) {
-        speex_encoder_destroy (enc->state);
-        enc->state = NULL;
-      }
-      speex_bits_destroy (&enc->bits);
-      break;
-    case GST_STATE_CHANGE_READY_TO_NULL:
-      gst_tag_list_free (enc->tags);
-      enc->tags = NULL;
-    default:
-      break;
-  }
-
-  return res;
-}
diff --git a/ext/speex/gstspeexenc.h b/ext/speex/gstspeexenc.h
index 6de9ebd45..17cb1e2df 100644
--- a/ext/speex/gstspeexenc.h
+++ b/ext/speex/gstspeexenc.h
@@ -23,7 +23,7 @@
 
 
 #include <gst/gst.h>
-#include <gst/base/gstadapter.h>
+#include <gst/audio/gstaudioencoder.h>
 
 #include <speex/speex.h>
 #include <speex/speex_header.h>
@@ -53,14 +53,7 @@ typedef struct _GstSpeexEnc GstSpeexEnc;
 typedef struct _GstSpeexEncClass GstSpeexEncClass;
 
 struct _GstSpeexEnc {
-  GstElement            element;
-
-  /* pads */
-  GstPad                *sinkpad,
-                        *srcpad;
-
-  gint                  packet_count;
-  gint                  n_packets;
+  GstAudioEncoder   element;
 
   SpeexBits             bits;
   SpeexHeader           header;
@@ -70,9 +63,9 @@ struct _GstSpeexEnc {
   const SpeexMode       *speex_mode;
 #endif
   void                  *state;
-  GstSpeexMode          mode;
-  GstAdapter            *adapter;
 
+  /* properties */
+  GstSpeexMode          mode;
   gfloat                quality;
   gint                  bitrate;
   gboolean              vbr;
@@ -81,40 +74,24 @@ struct _GstSpeexEnc {
   gboolean              dtx;
   gint                  complexity;
   gint                  nframes;
-
-  gint                  lookahead;
+  gchar                 *last_message;
 
   gint                  channels;
   gint                  rate;
 
-  gboolean              setup;
   gboolean              header_sent;
 
-  guint64               samples_in;
-  guint64               bytes_out;
-
   GstTagList            *tags;
 
-  gchar                 *last_message;
-
   gint                  frame_size;
-  guint64               frameno;
-  guint64               frameno_out;
+  gint                  lookahead;
 
   guint8                *comments;
   gint                  comment_len;
-
-  /* Timestamp and granulepos tracking */
-  GstClockTime     start_ts;
-  GstClockTime     next_ts;
-  guint64          granulepos_offset;
 };
 
 struct _GstSpeexEncClass {
-  GstElementClass parent_class;
-
-  /* signals */
-  void (*frame_encoded) (GstElement *element);
+  GstAudioEncoderClass parent_class;
 };
 
 GType gst_speex_enc_get_type (void);