modules/text_renderer/nsspeechsynthesizer.m

   1 /*****************************************************************************
   2  * nsspeechsynthesizer.m: Simple text to Speech renderer for Mac OS X
   3  *****************************************************************************
   4  * Copyright (C) 2015 VLC authors and VideoLAN
   5  * $Id$
   6  *
   7  * Authors: Felix Paul Kühne <fkuehne # videolan # org>
   8  *
   9  * This program is free software; you can redistribute it and/or modify it
  10  * under the terms of the GNU Lesser General Public License as published by
  11  * the Free Software Foundation; either version 2.1 of the License, or
  12  * (at your option) any later version.
  13  *
  14  * This program is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  17  * GNU Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public License
  20  * along with this program; if not, write to the Free Software Foundation,
  21  * Inc., 51 Franklin Street, Fifth Floor, Boston MA 02110-1301, USA.
  22  *****************************************************************************/
  23
  24 /*****************************************************************************
  25  * Preamble
  26  *****************************************************************************/
  27
  28 #ifdef HAVE_CONFIG_H
  29 # include "config.h"
  30 #endif
  31
  32 #include <vlc_common.h>
  33 #include <vlc_plugin.h>
  34 #include <vlc_filter.h>
  35 #include <vlc_subpicture.h>
  36
  37 #import <Cocoa/Cocoa.h>
  38
  39 static int Create (vlc_object_t *);
  40 static void Destroy(vlc_object_t *);
  41 static int RenderText(filter_t *,
  42                       subpicture_region_t *,
  43                       subpicture_region_t *,
  44                       const vlc_fourcc_t *);
  45
  46 vlc_module_begin ()
  47 set_description(N_("Speech synthesis for Mac OS X"))
  48 set_category(CAT_VIDEO)
  49 set_subcategory(SUBCAT_VIDEO_SUBPIC)
  50
  51 set_capability("text renderer", 0)
  52 set_callbacks(Create, Destroy)
  53 vlc_module_end ()
  54
  55 struct filter_sys_t
  56 {
  57     NSSpeechSynthesizer *speechSynthesizer;
  58     NSString *currentLocale;
  59     NSString *lastString;
  60 };
  61
  62 static int  Create (vlc_object_t *p_this)
  63 {
  64     filter_t *p_filter = (filter_t *)p_this;
  65     filter_sys_t *p_sys;
  66
  67     p_filter->p_sys = p_sys = malloc(sizeof(filter_sys_t));
  68     if (!p_sys)
  69         return VLC_ENOMEM;
  70
  71     p_sys->currentLocale = p_sys->lastString = @"";
  72     p_sys->speechSynthesizer = [[NSSpeechSynthesizer alloc] init];
  73
  74     p_filter->pf_render = RenderText;
  75
  76     return VLC_SUCCESS;
  77 }
  78
  79 static void Destroy(vlc_object_t *p_this)
  80 {
  81     filter_t *p_filter = (filter_t *)p_this;
  82     filter_sys_t *p_sys = p_filter->p_sys;
  83
  84     [p_sys->speechSynthesizer stopSpeaking];
  85     [p_sys->speechSynthesizer release];
  86     p_sys->speechSynthesizer = nil;
  87
  88     [p_sys->lastString release];
  89     p_sys->lastString = nil;
  90
  91     [p_sys->currentLocale release];
  92     p_sys->currentLocale = nil;
  93
  94     free(p_sys);
  95 }
  96
  97 static NSString * languageCodeForString(NSString *string) {
  98     return (NSString *)CFStringTokenizerCopyBestStringLanguage((CFStringRef)string, CFRangeMake(0, [string length]));
  99 }
 100
 101 static int RenderText(filter_t *p_filter,
 102                       subpicture_region_t *p_region_out,
 103                       subpicture_region_t *p_region_in,
 104                       const vlc_fourcc_t *p_chroma_list)
 105 {
 106     @autoreleasepool {
 107         filter_sys_t *p_sys = p_filter->p_sys;
 108         text_segment_t *p_segment = p_region_in->p_text;
 109
 110         if (!p_segment)
 111             return VLC_EGENERIC;
 112
 113         for ( const text_segment_t *s = p_segment; s != NULL; s = s->p_next ) {
 114             if ( !s->psz_text )
 115                 continue;
 116
 117             if (strlen(s->psz_text) == 0)
 118                 continue;
 119
 120             NSString *stringToSpeech = [NSString stringWithUTF8String:s->psz_text];
 121
 122             if ([p_sys->lastString isEqualToString:stringToSpeech])
 123                 continue;
 124
 125             if ([stringToSpeech isEqualToString:@"\n"])
 126                 continue;
 127
 128             p_sys->lastString = [stringToSpeech retain];
 129
 130             msg_Dbg(p_filter, "Speaking '%s'", [stringToSpeech UTF8String]);
 131
 132             NSString *detectedLocale = languageCodeForString(stringToSpeech);
 133             if (detectedLocale != nil) {
 134                 if (![detectedLocale isEqualToString:p_sys->currentLocale]) {
 135                     p_sys->currentLocale = [detectedLocale retain];
 136                     msg_Dbg(p_filter, "switching speaker locale to '%s'", [p_sys->currentLocale UTF8String]);
 137                     NSArray *voices = [NSSpeechSynthesizer availableVoices];
 138                     NSUInteger count = voices.count;
 139                     NSRange range = NSMakeRange(0, 2);
 140
 141                     for (NSUInteger i = 0; i < count; i++) {
 142                         NSDictionary *voiceAttributes = [NSSpeechSynthesizer attributesForVoice: [voices objectAtIndex:i]];
 143                         NSString *voiceLanguage = [voiceAttributes objectForKey:@"VoiceLanguage"];
 144                         if ([p_sys->currentLocale isEqualToString:[voiceLanguage substringWithRange:range]]) {
 145                             NSString *voiceName = [voiceAttributes objectForKey:@"VoiceName"];
 146                             msg_Dbg(p_filter, "switched to voice '%s'", [voiceName UTF8String]);
 147                             if ([voiceName isEqualToString:@"Agnes"] || [voiceName isEqualToString:@"Albert"])
 148                                 continue;
 149                             [p_sys->speechSynthesizer setVoice: [voices objectAtIndex:i]];
 150                             break;
 151                         }
 152                     }
 153                 }
 154             }
 155
 156             [p_sys->speechSynthesizer startSpeakingString:stringToSpeech];
 157         }
 158
 159         return VLC_SUCCESS;
 160     }
 161 }