Setup CAPS and do not use WAV format

4 files changed, 49 insertions, 43 deletions

diff --git a/src/espeak.c b/src/espeak.c
index dc8880b..265a243 100644
--- a/src/espeak.c
+++ b/src/espeak.c
@@ -27,24 +27,19 @@ struct Espeak
     GOutputStream *buffer;
 };
 
-static unsigned char wave_hdr[44] = {
-    'R','I','F','F',0x24,0xf0,0xff,0x7f,'W','A','V','E','f','m','t',' ',
-    0x10,0,0,0,1,0,1,0,  9,0x3d,0,0,0x12,0x7a,0,0,
-    2,0,0x10,0,'d','a','t','a',  0x00,0xf0,0xff,0x7f};
-
 pthread_mutex_t mutex = PTHREAD_MUTEX_INITIALIZER;
 static GOutputStream *buffer = NULL;
 static gint sample_rate = 0;
 static const espeak_VOICE **voices = NULL;
 
 static gint
-read_cb(short * wav, int numsamples, espeak_EVENT * events)
+read_cb(short * data, int numsamples, espeak_EVENT * events)
 {
-    if (wav == NULL)
+    if (data == NULL)
         return 0;
 
     if (numsamples > 0)
-        g_output_stream_write(buffer, wav, numsamples*2, NULL, NULL);
+        g_output_stream_write(buffer, data, numsamples*2, NULL, NULL);
 
     return 0;
 }
@@ -79,6 +74,12 @@ espeak_new()
     return es;
 }
 
+gint
+espeak_sample_rate()
+{
+    return sample_rate;
+}
+
 gchar**
 espeak_voices()
 {
@@ -101,9 +102,6 @@ espeak_say(struct Espeak *es, const gchar *text, const gchar *voice,
         guint pitch, guint rate)
 {
     g_seekable_seek(G_SEEKABLE(es->buffer), 0, G_SEEK_SET, NULL, NULL);
-    g_output_stream_write(es->buffer, wave_hdr, 24, NULL, NULL);
-    g_seekable_seek(G_SEEKABLE(es->buffer), 8, G_SEEK_CUR, NULL, NULL);
-    g_output_stream_write(es->buffer, wave_hdr+32, 12, NULL, NULL);
 
     pthread_mutex_lock(&mutex);
     buffer = es->buffer;
@@ -118,26 +116,6 @@ espeak_say(struct Espeak *es, const gchar *text, const gchar *voice,
     if (status != EE_OK)
         return FALSE;
 
-    void write4bytes(unsigned char *ptr, int value)
-    {
-        int ix;
-
-        for(ix=0; ix<4; ix++)
-        {
-            *ptr++ = value & 0xff;
-            value = value >> 8;
-        }
-    }
-
-    GMemoryOutputStream *mb = G_MEMORY_OUTPUT_STREAM(es->buffer);
-    unsigned char *ptr = g_memory_output_stream_get_data(mb);
-    guint size = g_memory_output_stream_get_data_size(mb);
-
-    write4bytes(ptr+24, sample_rate);
-    write4bytes(ptr+28, sample_rate*2);
-	write4bytes(ptr+4, size-8);
-	write4bytes(ptr+40, size-44);
-
     return TRUE;
 }
 
diff --git a/src/espeak.h b/src/espeak.h
index 8f7ef61..6258ebd 100644
--- a/src/espeak.h
+++ b/src/espeak.h
@@ -21,6 +21,7 @@
 struct Espeak;
 
 struct Espeak* espeak_new();
+gint espeak_sample_rate();
 gchar** espeak_voices();
 gboolean espeak_say(struct Espeak*, const gchar *text, const gchar *voice,
         guint pitch, guint rate);
diff --git a/src/gstespeak.c b/src/gstespeak.c
index 52cfacc..8fd36eb 100644
--- a/src/gstespeak.c
+++ b/src/gstespeak.c
@@ -23,7 +23,7 @@
  * <refsect2>
  * <title>Example launch line</title>
  * |[
- * gst-launch-0.10 espeak text="Hello world" pitch=99 rate=300 voice=default ! wavparse ! alsasink
+ * gst-launch-0.10 espeak text="Hello world" pitch=99 rate=300 voice=default ! alsasink
  * ]|
  * </refsect2>
  */
@@ -48,27 +48,28 @@ enum
     PROP_PITCH,
     PROP_RATE,
     PROP_VOICE,
-    PROP_VOICES
+    PROP_VOICES,
+    PROP_CAPS
 };
 
 static GstStaticPadTemplate src_factory = GST_STATIC_PAD_TEMPLATE (
     "src",
     GST_PAD_SRC,
     GST_PAD_ALWAYS,
-    GST_STATIC_CAPS ("ANY")
+    GST_STATIC_CAPS_ANY
     );
 
-static GstFlowReturn gst_espeak_create (GstBaseSrc*,
+static GstFlowReturn gst_espeak_create(GstBaseSrc*,
     guint64, guint, GstBuffer**);
-static gboolean gst_espeak_start (GstBaseSrc*);
-static gboolean gst_espeak_stop (GstBaseSrc*);
+static gboolean gst_espeak_start(GstBaseSrc*);
+static gboolean gst_espeak_stop(GstBaseSrc*);
 static gboolean gst_espeak_is_seekable (GstBaseSrc*);
 static void gst_espeak_init_uri(GType);
-static void gst_espeak_finalize(GObject * gobject);
-static void gst_espeak_set_property (GObject * object, guint prop_id,
-    const GValue * value, GParamSpec * pspec);
-static void gst_espeak_get_property (GObject * object, guint prop_id,
-    GValue * value, GParamSpec * pspec);
+static void gst_espeak_finalize(GObject*);
+static void gst_espeak_set_property(GObject*, guint, const GValue*,
+        GParamSpec*);
+static void gst_espeak_get_property(GObject*, guint, GValue*, GParamSpec*);
+static GstCaps *gst_espeak_getcaps(GstBaseSrc*);
 
 GST_BOILERPLATE_FULL(GstEspeak, gst_espeak, GstBaseSrc, GST_TYPE_BASE_SRC,
         gst_espeak_init_uri);
@@ -102,6 +103,7 @@ gst_espeak_class_init (GstEspeakClass * klass)
     basesrc_class->stop = gst_espeak_stop;
     basesrc_class->stop = gst_espeak_stop;
     basesrc_class->is_seekable = gst_espeak_is_seekable;
+    basesrc_class->get_caps = gst_espeak_getcaps;
 
     gobject_class->finalize = gst_espeak_finalize;
     gobject_class->set_property = gst_espeak_set_property;
@@ -127,6 +129,10 @@ gst_espeak_class_init (GstEspeakClass * klass)
             g_param_spec_boxed("voices", "List of voices",
                 "List of voices", G_TYPE_STRV,
                 G_PARAM_READABLE | G_PARAM_STATIC_STRINGS));
+    g_object_class_install_property (gobject_class, PROP_CAPS,
+            g_param_spec_boxed ("caps", "Caps",
+                "Caps describing the format of the data.", GST_TYPE_CAPS,
+                G_PARAM_READABLE));
 }
 
 /* initialize the new element
@@ -145,13 +151,23 @@ gst_espeak_init (GstEspeak * self,
     self->speak = espeak_new();
     self->voice = g_strdup("default");
     self->voices = espeak_voices();
+
+    self->caps = gst_caps_new_simple("audio/x-raw-int",
+            "rate", G_TYPE_INT, espeak_sample_rate(),
+            "channels", G_TYPE_INT, 1,
+            "endianness", G_TYPE_INT, G_BYTE_ORDER,
+            "width", G_TYPE_INT, 16,
+            "depth", G_TYPE_INT, 16,
+            "signed", G_TYPE_BOOLEAN, TRUE,
+            NULL);
 }
 
 static void
 gst_espeak_finalize(GObject * self_)
 {
-    GstEspeak *self = (GstEspeak*)self_;
+    GstEspeak *self = GST_ESPEAK(self_);
 
+    gst_caps_unref(self->caps); self->caps = NULL;
     espeak_unref(self->speak);  self->speak = NULL;
     g_free(self->text);         self->text = NULL;
     g_free(self->uri);          self->uri = NULL;
@@ -243,6 +259,9 @@ gst_espeak_get_property (GObject * object, guint prop_id,
         case PROP_VOICES:
             g_value_set_boxed(value, self->voices);
             break;
+        case PROP_CAPS:
+            gst_value_set_caps(value, self->caps);
+            break;
         default:
             G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
             break;
@@ -293,6 +312,13 @@ gst_espeak_is_seekable (GstBaseSrc * src)
     return FALSE;
 }
 
+static GstCaps *
+gst_espeak_getcaps(GstBaseSrc *self_)
+{
+    GstEspeak *self = GST_ESPEAK(self_);
+    return gst_caps_ref(self->caps);
+}
+
 /******************************************************************************/
 
 static GstURIType
diff --git a/src/gstespeak.h b/src/gstespeak.h
index 7af8b77..a0776aa 100644
--- a/src/gstespeak.h
+++ b/src/gstespeak.h
@@ -49,6 +49,7 @@ struct _GstEspeak
     guint rate;
     gchar *voice;
     gchar **voices;
+    GstCaps *caps;
 };
 
 struct _GstEspeakClass