1 /*
2 * Copyright (c) 2012 Stefano Sabatini
3 *
4 * This file is part of FFmpeg.
5 *
6 * FFmpeg is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
10 *
11 * FFmpeg is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
15 *
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 */
20
21 /**
22 * @file
23 * flite voice synth source
24 */
25
26 #include <flite/flite.h>
27 #include "libavutil/channel_layout.h"
28 #include "libavutil/file.h"
29 #include "libavutil/opt.h"
30 #include "libavutil/thread.h"
31 #include "avfilter.h"
32 #include "audio.h"
33 #include "formats.h"
34 #include "internal.h"
35
36 typedef struct FliteContext {
37 const AVClass *class;
38 char *voice_str;
39 char *textfile;
40 char *text;
41 cst_wave *wave;
42 int16_t *wave_samples;
43 int wave_nb_samples;
44 int list_voices;
45 cst_voice *voice;
46 struct voice_entry *voice_entry;
47 int64_t pts;
48 int frame_nb_samples; ///< number of samples per frame
49 } FliteContext;
50
51 #define OFFSET(x) offsetof(FliteContext, x)
52 #define FLAGS AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
53
54 static const AVOption flite_options[] = {
55 { "list_voices", "list voices and exit", OFFSET(list_voices), AV_OPT_TYPE_BOOL, {.i64=0}, 0, 1, FLAGS },
56 { "nb_samples", "set number of samples per frame", OFFSET(frame_nb_samples), AV_OPT_TYPE_INT, {.i64=512}, 0, INT_MAX, FLAGS },
57 { "n", "set number of samples per frame", OFFSET(frame_nb_samples), AV_OPT_TYPE_INT, {.i64=512}, 0, INT_MAX, FLAGS },
58 { "text", "set text to speak", OFFSET(text), AV_OPT_TYPE_STRING, {.str=NULL}, 0, 0, FLAGS },
59 { "textfile", "set filename of the text to speak", OFFSET(textfile), AV_OPT_TYPE_STRING, {.str=NULL}, 0, 0, FLAGS },
60 { "v", "set voice", OFFSET(voice_str), AV_OPT_TYPE_STRING, {.str="kal"}, 0, 0, FLAGS },
61 { "voice", "set voice", OFFSET(voice_str), AV_OPT_TYPE_STRING, {.str="kal"}, 0, 0, FLAGS },
62 { NULL }
63 };
64
65 AVFILTER_DEFINE_CLASS(flite);
66
67 static AVMutex flite_mutex = AV_MUTEX_INITIALIZER;
68
69 static int flite_inited = 0;
70
71 /* declare functions for all the supported voices */
72 #define DECLARE_REGISTER_VOICE_FN(name) \
73 cst_voice *register_cmu_us_## name(const char *); \
74 void unregister_cmu_us_## name(cst_voice *)
75 DECLARE_REGISTER_VOICE_FN(awb);
76 DECLARE_REGISTER_VOICE_FN(kal);
77 DECLARE_REGISTER_VOICE_FN(kal16);
78 DECLARE_REGISTER_VOICE_FN(rms);
79 DECLARE_REGISTER_VOICE_FN(slt);
80
81 struct voice_entry {
82 const char *name;
83 cst_voice * (*register_fn)(const char *);
84 void (*unregister_fn)(cst_voice *);
85 cst_voice *voice;
86 unsigned usage_count;
87 };
88
89 #define MAKE_VOICE_STRUCTURE(voice_name) { \
90 .name = #voice_name, \
91 .register_fn = register_cmu_us_ ## voice_name, \
92 .unregister_fn = unregister_cmu_us_ ## voice_name, \
93 }
94 static struct voice_entry voice_entries[] = {
95 MAKE_VOICE_STRUCTURE(awb),
96 MAKE_VOICE_STRUCTURE(kal),
97 MAKE_VOICE_STRUCTURE(kal16),
98 MAKE_VOICE_STRUCTURE(rms),
99 MAKE_VOICE_STRUCTURE(slt),
100 };
101
list_voices(void * log_ctx,const char * sep)102 static void list_voices(void *log_ctx, const char *sep)
103 {
104 int i, n = FF_ARRAY_ELEMS(voice_entries);
105 for (i = 0; i < n; i++)
106 av_log(log_ctx, AV_LOG_INFO, "%s%s",
107 voice_entries[i].name, i < (n-1) ? sep : "\n");
108 }
109
select_voice(struct voice_entry ** entry_ret,const char * voice_name,void * log_ctx)110 static int select_voice(struct voice_entry **entry_ret, const char *voice_name, void *log_ctx)
111 {
112 int i;
113
114 for (i = 0; i < FF_ARRAY_ELEMS(voice_entries); i++) {
115 struct voice_entry *entry = &voice_entries[i];
116 if (!strcmp(entry->name, voice_name)) {
117 cst_voice *voice;
118 pthread_mutex_lock(&flite_mutex);
119 if (!entry->voice)
120 entry->voice = entry->register_fn(NULL);
121 voice = entry->voice;
122 if (voice)
123 entry->usage_count++;
124 pthread_mutex_unlock(&flite_mutex);
125 if (!voice) {
126 av_log(log_ctx, AV_LOG_ERROR,
127 "Could not register voice '%s'\n", voice_name);
128 return AVERROR_UNKNOWN;
129 }
130 *entry_ret = entry;
131 return 0;
132 }
133 }
134
135 av_log(log_ctx, AV_LOG_ERROR, "Could not find voice '%s'\n", voice_name);
136 av_log(log_ctx, AV_LOG_INFO, "Choose between the voices: ");
137 list_voices(log_ctx, ", ");
138
139 return AVERROR(EINVAL);
140 }
141
init(AVFilterContext * ctx)142 static av_cold int init(AVFilterContext *ctx)
143 {
144 FliteContext *flite = ctx->priv;
145 int ret = 0;
146
147 if (flite->list_voices) {
148 list_voices(ctx, "\n");
149 return AVERROR_EXIT;
150 }
151
152 pthread_mutex_lock(&flite_mutex);
153 if (!flite_inited) {
154 if ((ret = flite_init()) >= 0)
155 flite_inited = 1;
156 }
157 pthread_mutex_unlock(&flite_mutex);
158 if (ret < 0) {
159 av_log(ctx, AV_LOG_ERROR, "flite initialization failed\n");
160 return AVERROR_EXTERNAL;
161 }
162
163 if ((ret = select_voice(&flite->voice_entry, flite->voice_str, ctx)) < 0)
164 return ret;
165 flite->voice = flite->voice_entry->voice;
166
167 if (flite->textfile && flite->text) {
168 av_log(ctx, AV_LOG_ERROR,
169 "Both text and textfile options set: only one must be specified\n");
170 return AVERROR(EINVAL);
171 }
172
173 if (flite->textfile) {
174 uint8_t *textbuf;
175 size_t textbuf_size;
176
177 if ((ret = av_file_map(flite->textfile, &textbuf, &textbuf_size, 0, ctx)) < 0) {
178 av_log(ctx, AV_LOG_ERROR,
179 "The text file '%s' could not be read: %s\n",
180 flite->textfile, av_err2str(ret));
181 return ret;
182 }
183
184 if (!(flite->text = av_malloc(textbuf_size+1))) {
185 av_file_unmap(textbuf, textbuf_size);
186 return AVERROR(ENOMEM);
187 }
188 memcpy(flite->text, textbuf, textbuf_size);
189 flite->text[textbuf_size] = 0;
190 av_file_unmap(textbuf, textbuf_size);
191 }
192
193 if (!flite->text) {
194 av_log(ctx, AV_LOG_ERROR,
195 "No speech text specified, specify the 'text' or 'textfile' option\n");
196 return AVERROR(EINVAL);
197 }
198
199 /* synth all the file data in block */
200 flite->wave = flite_text_to_wave(flite->text, flite->voice);
201 flite->wave_samples = flite->wave->samples;
202 flite->wave_nb_samples = flite->wave->num_samples;
203 return 0;
204 }
205
uninit(AVFilterContext * ctx)206 static av_cold void uninit(AVFilterContext *ctx)
207 {
208 FliteContext *flite = ctx->priv;
209
210 if (flite->voice_entry) {
211 pthread_mutex_lock(&flite_mutex);
212 if (!--flite->voice_entry->usage_count) {
213 flite->voice_entry->unregister_fn(flite->voice);
214 flite->voice_entry->voice = NULL;
215 }
216 pthread_mutex_unlock(&flite_mutex);
217 }
218 delete_wave(flite->wave);
219 flite->wave = NULL;
220 }
221
query_formats(AVFilterContext * ctx)222 static int query_formats(AVFilterContext *ctx)
223 {
224 FliteContext *flite = ctx->priv;
225 int ret;
226
227 AVFilterChannelLayouts *chlayouts = NULL;
228 AVFilterFormats *sample_formats = NULL;
229 AVFilterFormats *sample_rates = NULL;
230 AVChannelLayout chlayout = { 0 };
231
232 av_channel_layout_default(&chlayout, flite->wave->num_channels);
233
234 if ((ret = ff_add_channel_layout (&chlayouts , &chlayout )) < 0 ||
235 (ret = ff_set_common_channel_layouts (ctx , chlayouts )) < 0 ||
236 (ret = ff_add_format (&sample_formats, AV_SAMPLE_FMT_S16 )) < 0 ||
237 (ret = ff_set_common_formats (ctx , sample_formats )) < 0 ||
238 (ret = ff_add_format (&sample_rates , flite->wave->sample_rate)) < 0 ||
239 (ret = ff_set_common_samplerates (ctx , sample_rates )) < 0)
240 return ret;
241
242 return 0;
243 }
244
config_props(AVFilterLink * outlink)245 static int config_props(AVFilterLink *outlink)
246 {
247 AVFilterContext *ctx = outlink->src;
248 FliteContext *flite = ctx->priv;
249
250 outlink->sample_rate = flite->wave->sample_rate;
251 outlink->time_base = (AVRational){1, flite->wave->sample_rate};
252
253 av_log(ctx, AV_LOG_VERBOSE, "voice:%s fmt:%s sample_rate:%d\n",
254 flite->voice_str,
255 av_get_sample_fmt_name(outlink->format), outlink->sample_rate);
256 return 0;
257 }
258
request_frame(AVFilterLink * outlink)259 static int request_frame(AVFilterLink *outlink)
260 {
261 AVFrame *samplesref;
262 FliteContext *flite = outlink->src->priv;
263 int nb_samples = FFMIN(flite->wave_nb_samples, flite->frame_nb_samples);
264
265 if (!nb_samples)
266 return AVERROR_EOF;
267
268 samplesref = ff_get_audio_buffer(outlink, nb_samples);
269 if (!samplesref)
270 return AVERROR(ENOMEM);
271
272 memcpy(samplesref->data[0], flite->wave_samples,
273 nb_samples * flite->wave->num_channels * 2);
274 samplesref->pts = flite->pts;
275 samplesref->pkt_pos = -1;
276 samplesref->sample_rate = flite->wave->sample_rate;
277 flite->pts += nb_samples;
278 flite->wave_samples += nb_samples * flite->wave->num_channels;
279 flite->wave_nb_samples -= nb_samples;
280
281 return ff_filter_frame(outlink, samplesref);
282 }
283
284 static const AVFilterPad flite_outputs[] = {
285 {
286 .name = "default",
287 .type = AVMEDIA_TYPE_AUDIO,
288 .config_props = config_props,
289 .request_frame = request_frame,
290 },
291 };
292
293 const AVFilter ff_asrc_flite = {
294 .name = "flite",
295 .description = NULL_IF_CONFIG_SMALL("Synthesize voice from text using libflite."),
296 .init = init,
297 .uninit = uninit,
298 .priv_size = sizeof(FliteContext),
299 .inputs = NULL,
300 FILTER_OUTPUTS(flite_outputs),
301 FILTER_QUERY_FUNC(query_formats),
302 .priv_class = &flite_class,
303 };
304