| 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 "avfilter.h" |
| 31 | #include "audio.h" |
| 32 | #include "formats.h" |
| 33 | #include "internal.h" |
| 34 | |
| 35 | typedef struct { |
| 36 | const AVClass *class; |
| 37 | char *voice_str; |
| 38 | char *textfile; |
| 39 | char *text; |
| 40 | cst_wave *wave; |
| 41 | int16_t *wave_samples; |
| 42 | int wave_nb_samples; |
| 43 | int list_voices; |
| 44 | cst_voice *voice; |
| 45 | struct voice_entry *voice_entry; |
| 46 | int64_t pts; |
| 47 | int frame_nb_samples; ///< number of samples per frame |
| 48 | } FliteContext; |
| 49 | |
| 50 | #define OFFSET(x) offsetof(FliteContext, x) |
| 51 | #define FLAGS AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM |
| 52 | |
| 53 | static const AVOption flite_options[] = { |
| 54 | { "list_voices", "list voices and exit", OFFSET(list_voices), AV_OPT_TYPE_INT, {.i64=0}, 0, 1, FLAGS }, |
| 55 | { "nb_samples", "set number of samples per frame", OFFSET(frame_nb_samples), AV_OPT_TYPE_INT, {.i64=512}, 0, INT_MAX, FLAGS }, |
| 56 | { "n", "set number of samples per frame", OFFSET(frame_nb_samples), AV_OPT_TYPE_INT, {.i64=512}, 0, INT_MAX, FLAGS }, |
| 57 | { "text", "set text to speak", OFFSET(text), AV_OPT_TYPE_STRING, {.str=NULL}, CHAR_MIN, CHAR_MAX, FLAGS }, |
| 58 | { "textfile", "set filename of the text to speak", OFFSET(textfile), AV_OPT_TYPE_STRING, {.str=NULL}, CHAR_MIN, CHAR_MAX, FLAGS }, |
| 59 | { "v", "set voice", OFFSET(voice_str), AV_OPT_TYPE_STRING, {.str="kal"}, CHAR_MIN, CHAR_MAX, FLAGS }, |
| 60 | { "voice", "set voice", OFFSET(voice_str), AV_OPT_TYPE_STRING, {.str="kal"}, CHAR_MIN, CHAR_MAX, FLAGS }, |
| 61 | { NULL } |
| 62 | }; |
| 63 | |
| 64 | AVFILTER_DEFINE_CLASS(flite); |
| 65 | |
| 66 | static volatile int flite_inited = 0; |
| 67 | |
| 68 | /* declare functions for all the supported voices */ |
| 69 | #define DECLARE_REGISTER_VOICE_FN(name) \ |
| 70 | cst_voice *register_cmu_us_## name(const char *); \ |
| 71 | void unregister_cmu_us_## name(cst_voice *); |
| 72 | DECLARE_REGISTER_VOICE_FN(awb); |
| 73 | DECLARE_REGISTER_VOICE_FN(kal); |
| 74 | DECLARE_REGISTER_VOICE_FN(kal16); |
| 75 | DECLARE_REGISTER_VOICE_FN(rms); |
| 76 | DECLARE_REGISTER_VOICE_FN(slt); |
| 77 | |
| 78 | struct voice_entry { |
| 79 | const char *name; |
| 80 | cst_voice * (*register_fn)(const char *); |
| 81 | void (*unregister_fn)(cst_voice *); |
| 82 | cst_voice *voice; |
| 83 | unsigned usage_count; |
| 84 | } voice_entry; |
| 85 | |
| 86 | #define MAKE_VOICE_STRUCTURE(voice_name) { \ |
| 87 | .name = #voice_name, \ |
| 88 | .register_fn = register_cmu_us_ ## voice_name, \ |
| 89 | .unregister_fn = unregister_cmu_us_ ## voice_name, \ |
| 90 | } |
| 91 | static struct voice_entry voice_entries[] = { |
| 92 | MAKE_VOICE_STRUCTURE(awb), |
| 93 | MAKE_VOICE_STRUCTURE(kal), |
| 94 | MAKE_VOICE_STRUCTURE(kal16), |
| 95 | MAKE_VOICE_STRUCTURE(rms), |
| 96 | MAKE_VOICE_STRUCTURE(slt), |
| 97 | }; |
| 98 | |
| 99 | static void list_voices(void *log_ctx, const char *sep) |
| 100 | { |
| 101 | int i, n = FF_ARRAY_ELEMS(voice_entries); |
| 102 | for (i = 0; i < n; i++) |
| 103 | av_log(log_ctx, AV_LOG_INFO, "%s%s", |
| 104 | voice_entries[i].name, i < (n-1) ? sep : "\n"); |
| 105 | } |
| 106 | |
| 107 | static int select_voice(struct voice_entry **entry_ret, const char *voice_name, void *log_ctx) |
| 108 | { |
| 109 | int i; |
| 110 | |
| 111 | for (i = 0; i < FF_ARRAY_ELEMS(voice_entries); i++) { |
| 112 | struct voice_entry *entry = &voice_entries[i]; |
| 113 | if (!strcmp(entry->name, voice_name)) { |
| 114 | if (!entry->voice) |
| 115 | entry->voice = entry->register_fn(NULL); |
| 116 | if (!entry->voice) { |
| 117 | av_log(log_ctx, AV_LOG_ERROR, |
| 118 | "Could not register voice '%s'\n", voice_name); |
| 119 | return AVERROR_UNKNOWN; |
| 120 | } |
| 121 | entry->usage_count++; |
| 122 | *entry_ret = entry; |
| 123 | return 0; |
| 124 | } |
| 125 | } |
| 126 | |
| 127 | av_log(log_ctx, AV_LOG_ERROR, "Could not find voice '%s'\n", voice_name); |
| 128 | av_log(log_ctx, AV_LOG_INFO, "Choose between the voices: "); |
| 129 | list_voices(log_ctx, ", "); |
| 130 | |
| 131 | return AVERROR(EINVAL); |
| 132 | } |
| 133 | |
| 134 | static av_cold int init(AVFilterContext *ctx) |
| 135 | { |
| 136 | FliteContext *flite = ctx->priv; |
| 137 | int ret = 0; |
| 138 | |
| 139 | if (flite->list_voices) { |
| 140 | list_voices(ctx, "\n"); |
| 141 | return AVERROR_EXIT; |
| 142 | } |
| 143 | |
| 144 | if (!flite_inited) { |
| 145 | if (flite_init() < 0) { |
| 146 | av_log(ctx, AV_LOG_ERROR, "flite initialization failed\n"); |
| 147 | return AVERROR_UNKNOWN; |
| 148 | } |
| 149 | flite_inited++; |
| 150 | } |
| 151 | |
| 152 | if ((ret = select_voice(&flite->voice_entry, flite->voice_str, ctx)) < 0) |
| 153 | return ret; |
| 154 | flite->voice = flite->voice_entry->voice; |
| 155 | |
| 156 | if (flite->textfile && flite->text) { |
| 157 | av_log(ctx, AV_LOG_ERROR, |
| 158 | "Both text and textfile options set: only one must be specified\n"); |
| 159 | return AVERROR(EINVAL); |
| 160 | } |
| 161 | |
| 162 | if (flite->textfile) { |
| 163 | uint8_t *textbuf; |
| 164 | size_t textbuf_size; |
| 165 | |
| 166 | if ((ret = av_file_map(flite->textfile, &textbuf, &textbuf_size, 0, ctx)) < 0) { |
| 167 | av_log(ctx, AV_LOG_ERROR, |
| 168 | "The text file '%s' could not be read: %s\n", |
| 169 | flite->textfile, av_err2str(ret)); |
| 170 | return ret; |
| 171 | } |
| 172 | |
| 173 | if (!(flite->text = av_malloc(textbuf_size+1))) |
| 174 | return AVERROR(ENOMEM); |
| 175 | memcpy(flite->text, textbuf, textbuf_size); |
| 176 | flite->text[textbuf_size] = 0; |
| 177 | av_file_unmap(textbuf, textbuf_size); |
| 178 | } |
| 179 | |
| 180 | if (!flite->text) { |
| 181 | av_log(ctx, AV_LOG_ERROR, |
| 182 | "No speech text specified, specify the 'text' or 'textfile' option\n"); |
| 183 | return AVERROR(EINVAL); |
| 184 | } |
| 185 | |
| 186 | /* synth all the file data in block */ |
| 187 | flite->wave = flite_text_to_wave(flite->text, flite->voice); |
| 188 | flite->wave_samples = flite->wave->samples; |
| 189 | flite->wave_nb_samples = flite->wave->num_samples; |
| 190 | return 0; |
| 191 | } |
| 192 | |
| 193 | static av_cold void uninit(AVFilterContext *ctx) |
| 194 | { |
| 195 | FliteContext *flite = ctx->priv; |
| 196 | |
| 197 | if (!--flite->voice_entry->usage_count) |
| 198 | flite->voice_entry->unregister_fn(flite->voice); |
| 199 | flite->voice = NULL; |
| 200 | flite->voice_entry = NULL; |
| 201 | delete_wave(flite->wave); |
| 202 | flite->wave = NULL; |
| 203 | } |
| 204 | |
| 205 | static int query_formats(AVFilterContext *ctx) |
| 206 | { |
| 207 | FliteContext *flite = ctx->priv; |
| 208 | |
| 209 | AVFilterChannelLayouts *chlayouts = NULL; |
| 210 | int64_t chlayout = av_get_default_channel_layout(flite->wave->num_channels); |
| 211 | AVFilterFormats *sample_formats = NULL; |
| 212 | AVFilterFormats *sample_rates = NULL; |
| 213 | |
| 214 | ff_add_channel_layout(&chlayouts, chlayout); |
| 215 | ff_set_common_channel_layouts(ctx, chlayouts); |
| 216 | ff_add_format(&sample_formats, AV_SAMPLE_FMT_S16); |
| 217 | ff_set_common_formats(ctx, sample_formats); |
| 218 | ff_add_format(&sample_rates, flite->wave->sample_rate); |
| 219 | ff_set_common_samplerates (ctx, sample_rates); |
| 220 | |
| 221 | return 0; |
| 222 | } |
| 223 | |
| 224 | static int config_props(AVFilterLink *outlink) |
| 225 | { |
| 226 | AVFilterContext *ctx = outlink->src; |
| 227 | FliteContext *flite = ctx->priv; |
| 228 | |
| 229 | outlink->sample_rate = flite->wave->sample_rate; |
| 230 | outlink->time_base = (AVRational){1, flite->wave->sample_rate}; |
| 231 | |
| 232 | av_log(ctx, AV_LOG_VERBOSE, "voice:%s fmt:%s sample_rate:%d\n", |
| 233 | flite->voice_str, |
| 234 | av_get_sample_fmt_name(outlink->format), outlink->sample_rate); |
| 235 | return 0; |
| 236 | } |
| 237 | |
| 238 | static int request_frame(AVFilterLink *outlink) |
| 239 | { |
| 240 | AVFrame *samplesref; |
| 241 | FliteContext *flite = outlink->src->priv; |
| 242 | int nb_samples = FFMIN(flite->wave_nb_samples, flite->frame_nb_samples); |
| 243 | |
| 244 | if (!nb_samples) |
| 245 | return AVERROR_EOF; |
| 246 | |
| 247 | samplesref = ff_get_audio_buffer(outlink, nb_samples); |
| 248 | if (!samplesref) |
| 249 | return AVERROR(ENOMEM); |
| 250 | |
| 251 | memcpy(samplesref->data[0], flite->wave_samples, |
| 252 | nb_samples * flite->wave->num_channels * 2); |
| 253 | samplesref->pts = flite->pts; |
| 254 | av_frame_set_pkt_pos(samplesref, -1); |
| 255 | av_frame_set_sample_rate(samplesref, flite->wave->sample_rate); |
| 256 | flite->pts += nb_samples; |
| 257 | flite->wave_samples += nb_samples * flite->wave->num_channels; |
| 258 | flite->wave_nb_samples -= nb_samples; |
| 259 | |
| 260 | return ff_filter_frame(outlink, samplesref); |
| 261 | } |
| 262 | |
| 263 | static const AVFilterPad flite_outputs[] = { |
| 264 | { |
| 265 | .name = "default", |
| 266 | .type = AVMEDIA_TYPE_AUDIO, |
| 267 | .config_props = config_props, |
| 268 | .request_frame = request_frame, |
| 269 | }, |
| 270 | { NULL } |
| 271 | }; |
| 272 | |
| 273 | AVFilter ff_asrc_flite = { |
| 274 | .name = "flite", |
| 275 | .description = NULL_IF_CONFIG_SMALL("Synthesize voice from text using libflite."), |
| 276 | .query_formats = query_formats, |
| 277 | .init = init, |
| 278 | .uninit = uninit, |
| 279 | .priv_size = sizeof(FliteContext), |
| 280 | .inputs = NULL, |
| 281 | .outputs = flite_outputs, |
| 282 | .priv_class = &flite_class, |
| 283 | }; |