Hello,
A couple of days ago I posted the (attached) patch for gnome-speech that
solves the problems with the Italian synthesis.
The patch follows the discussion we had in the list:
- it asks the Festival voices what encoding they prefer as was
suggested by Milan Zamaza[1]
- it defaults to ISO-8859-1
I also tried to make the patch as unintrusive as possible: I choose to
maintain a static, private list of per-voice preferred encodings, so
that I didn't need to change any .h file, nor the gnome-speech IDL, to
store the encoding information. While there can probably be better
ways, this one should add the functionality without breaking anything
code-wise.
What this patch could break is the Indian festival voices, which want
UTF-8 on input and probably don't report the coding in their
(voice.description). I volunteer to prepare a patch for the Indian
festival voices to fix this, if someone tells me where they are.
Now, in order to carry on my work towards a speaking Italian Ubuntu
Dapper Gnome desktop, I'd like this patch to be accepted into
gnome-speech.
Once the patch has some official blessing, I can probably prod the Ubuntu
maintainers to release a gnome-speech update which would allow the
Italian desktop to speak on Dapper almost out of the box.
Please?
Ciao,
Enrico kneeling and patiently waiting for the spray of holy water.
[1] I'd love to post links to the list archives, but they seem not to be
updated.
--
GPG key: 1024D/797EBFAB 2000-12-05 Enrico Zini <enrico debian org>
--- festivalsynthesisdriver.c 2006-05-14 15:49:21.000000000 +0100
+++ festivalsynthesisdriver.c.buono 2006-06-29 17:23:25.000000000 +0100
@@ -45,6 +45,8 @@
#undef FESTIVAL_DEBUG_SEND
#undef FESTIVAL_DEBUG_TEXT
+#define DEFAULT_ENCODING "ISO-8859-1"
+
static gint text_id = 0;
static GObjectClass *parent_class;
static gboolean festival_server_exists = FALSE;
@@ -52,6 +54,7 @@
static GSList *driver_list = NULL;
static GSList *markers_list = NULL;
static GSList *voices_list = NULL;
+static GSList *encodings_list = NULL;
typedef struct
{
@@ -98,6 +101,37 @@
static void festival_process_text_out (FestivalTextOut *text_out);
static void festival_free_list (FestivalSynthesisDriver *d);
+static void encoding_set(GNOME_Speech_VoiceInfo* voice, gchar* encoding)
+{
+ /* First, try to see if we have it already */
+ GSList* cur = encodings_list;
+ int len = strlen(voice->name);
+ for ( ; cur != NULL; cur = cur->next)
+ if (strncmp((gchar*)cur->data, voice->name, len) == 0)
+ {
+ /* If we have it, replace it */
+ g_free(cur->data);
+ cur->data = g_strdup_printf("%s:%s", voice->name, encoding);
+ return;
+ }
+ /* Otherwise, insert it */
+ encodings_list = g_slist_prepend (encodings_list, g_strdup_printf("%s:%s", voice->name, encoding));
+}
+
+static gchar* encoding_get(FestivalSpeaker* speaker)
+{
+ /* Look for the item in the list */
+ GSList* cur = encodings_list;
+ int len = strlen(speaker->voice);
+ /* Speaker should be at least "(voice_?)\n" */
+ if (len < 10)
+ return DEFAULT_ENCODING;
+ for ( ; cur != NULL; cur = cur->next)
+ if (strncmp((gchar*)cur->data, speaker->voice + 7, len - 9) == 0)
+ return (gchar*)cur->data + len - 9 + 1;
+ /* If not found, default to latin1 */
+ return DEFAULT_ENCODING;
+}
static FestivalTextMarker *
festival_text_marker_new ()
@@ -329,12 +363,19 @@
p = strstr (voices[i], ")\n");
if (p != NULL)
*p = 0;
- query_voice_des = g_strdup_printf ("(list 'VOICEDESC: "
- "(nth 0 (voice.description '%s)) "
- "(nth 1 (nth 0 (nth 1 (voice.description '%s)))) "
- "(nth 1 (nth 2 (nth 1 (voice.description '%s)))) "
- "(nth 1 (nth 1 (nth 1 (voice.description '%s)))))\n",
- voices[i], voices[i], voices[i], voices[i]);
+ /* Adds an extra nil to help tokenization later */
+ query_voice_des = g_strdup_printf (
+ "(list 'VOICEDESC: "
+ "(nth 0 (voice.description '%s)) "
+ "(cadr (assoc 'language (cadr (voice.description '%s)))) "
+ "(cadr (assoc 'dialect (cadr (voice.description '%s)))) "
+ "(cadr (assoc 'gender (cadr (voice.description '%s)))) "
+ "(or "
+ " (cadr (assoc 'coding (cadr (voice.description '%s)))) "
+ " \"" DEFAULT_ENCODING "\") "
+ "nil)\n",
+ voices[i], voices[i], voices[i], voices[i], voices[i]);
+
festival_synthesis_driver_say_raw (driver, query_voice_des);
voices_waiting_for_description++;
g_free (query_voice_des);
@@ -349,7 +390,8 @@
voice[1] != NULL && strcmp (voice[1], "nil") &&
voice[2] != NULL && strcmp (voice[2], "nil") &&
voice[3] != NULL && strcmp (voice[3], "nil") &&
- voice[4] != NULL && strcmp (voice[4], "nil"))
+ voice[4] != NULL && strcmp (voice[4], "nil") &&
+ voice[5] != NULL && strcmp (voice[5], "nil"))
{
GNOME_Speech_VoiceInfo *new_info;
new_info = GNOME_Speech_VoiceInfo__alloc ();
@@ -360,6 +402,7 @@
else
new_info->gender = GNOME_Speech_gender_male;
voices_list = g_slist_append (voices_list, new_info);
+ encoding_set(new_info, voice[5]);
}
voices_waiting_for_description--;
if (voices_waiting_for_description == 0)
@@ -932,6 +975,7 @@
festival_synthesis_driver_say_raw (d, s->voice);
speaker_refresh_parameters (SPEAKER(s));
d->last_speaker = s;
+ g_io_channel_set_encoding(d->channel_sock, encoding_get(d->last_speaker), NULL);
}
clb_list_free (d->crt_clbs);
Attachment:
signature.asc
Description: Digital signature