-
Notifications
You must be signed in to change notification settings - Fork 503
/
piper.cpp
636 lines (519 loc) · 21.4 KB
/
piper.cpp
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
#include <array>
#include <chrono>
#include <fstream>
#include <limits>
#include <sstream>
#include <stdexcept>
#include <espeak-ng/speak_lib.h>
#include <onnxruntime_cxx_api.h>
#include <spdlog/spdlog.h>
#include "json.hpp"
#include "piper.hpp"
#include "utf8.h"
#include "wavfile.hpp"
namespace piper {
#ifdef _PIPER_VERSION
// https://stackoverflow.com/questions/47346133/how-to-use-a-define-inside-a-format-string
#define _STR(x) #x
#define STR(x) _STR(x)
const std::string VERSION = STR(_PIPER_VERSION);
#else
const std::string VERSION = "";
#endif
// Maximum value for 16-bit signed WAV sample
const float MAX_WAV_VALUE = 32767.0f;
const std::string instanceName{"piper"};
std::string getVersion() { return VERSION; }
// True if the string is a single UTF-8 codepoint
bool isSingleCodepoint(std::string s) {
return utf8::distance(s.begin(), s.end()) == 1;
}
// Get the first UTF-8 codepoint of a string
Phoneme getCodepoint(std::string s) {
utf8::iterator character_iter(s.begin(), s.begin(), s.end());
return *character_iter;
}
// Load JSON config information for phonemization
void parsePhonemizeConfig(json &configRoot, PhonemizeConfig &phonemizeConfig) {
// {
// "espeak": {
// "voice": "<language code>"
// },
// "phoneme_type": "<espeak or text>",
// "phoneme_map": {
// "<from phoneme>": ["<to phoneme 1>", "<to phoneme 2>", ...]
// },
// "phoneme_id_map": {
// "<phoneme>": [<id1>, <id2>, ...]
// }
// }
if (configRoot.contains("espeak")) {
auto espeakValue = configRoot["espeak"];
if (espeakValue.contains("voice")) {
phonemizeConfig.eSpeak.voice = espeakValue["voice"].get<std::string>();
}
}
if (configRoot.contains("phoneme_type")) {
auto phonemeTypeStr = configRoot["phoneme_type"].get<std::string>();
if (phonemeTypeStr == "text") {
phonemizeConfig.phonemeType = TextPhonemes;
}
}
// phoneme to [id] map
// Maps phonemes to one or more phoneme ids (required).
if (configRoot.contains("phoneme_id_map")) {
auto phonemeIdMapValue = configRoot["phoneme_id_map"];
for (auto &fromPhonemeItem : phonemeIdMapValue.items()) {
std::string fromPhoneme = fromPhonemeItem.key();
if (!isSingleCodepoint(fromPhoneme)) {
std::stringstream idsStr;
for (auto &toIdValue : fromPhonemeItem.value()) {
PhonemeId toId = toIdValue.get<PhonemeId>();
idsStr << toId << ",";
}
spdlog::error("\"{}\" is not a single codepoint (ids={})", fromPhoneme,
idsStr.str());
throw std::runtime_error(
"Phonemes must be one codepoint (phoneme id map)");
}
auto fromCodepoint = getCodepoint(fromPhoneme);
for (auto &toIdValue : fromPhonemeItem.value()) {
PhonemeId toId = toIdValue.get<PhonemeId>();
phonemizeConfig.phonemeIdMap[fromCodepoint].push_back(toId);
}
}
}
// phoneme to [phoneme] map
// Maps phonemes to one or more other phonemes (not normally used).
if (configRoot.contains("phoneme_map")) {
if (!phonemizeConfig.phonemeMap) {
phonemizeConfig.phonemeMap.emplace();
}
auto phonemeMapValue = configRoot["phoneme_map"];
for (auto &fromPhonemeItem : phonemeMapValue.items()) {
std::string fromPhoneme = fromPhonemeItem.key();
if (!isSingleCodepoint(fromPhoneme)) {
spdlog::error("\"{}\" is not a single codepoint", fromPhoneme);
throw std::runtime_error(
"Phonemes must be one codepoint (phoneme map)");
}
auto fromCodepoint = getCodepoint(fromPhoneme);
for (auto &toPhonemeValue : fromPhonemeItem.value()) {
std::string toPhoneme = toPhonemeValue.get<std::string>();
if (!isSingleCodepoint(toPhoneme)) {
throw std::runtime_error(
"Phonemes must be one codepoint (phoneme map)");
}
auto toCodepoint = getCodepoint(toPhoneme);
(*phonemizeConfig.phonemeMap)[fromCodepoint].push_back(toCodepoint);
}
}
}
} /* parsePhonemizeConfig */
// Load JSON config for audio synthesis
void parseSynthesisConfig(json &configRoot, SynthesisConfig &synthesisConfig) {
// {
// "audio": {
// "sample_rate": 22050
// },
// "inference": {
// "noise_scale": 0.667,
// "length_scale": 1,
// "noise_w": 0.8,
// "phoneme_silence": {
// "<phoneme>": <seconds of silence>,
// ...
// }
// }
// }
if (configRoot.contains("audio")) {
auto audioValue = configRoot["audio"];
if (audioValue.contains("sample_rate")) {
// Default sample rate is 22050 Hz
synthesisConfig.sampleRate = audioValue.value("sample_rate", 22050);
}
}
if (configRoot.contains("inference")) {
// Overrides default inference settings
auto inferenceValue = configRoot["inference"];
if (inferenceValue.contains("noise_scale")) {
synthesisConfig.noiseScale = inferenceValue.value("noise_scale", 0.667f);
}
if (inferenceValue.contains("length_scale")) {
synthesisConfig.lengthScale = inferenceValue.value("length_scale", 1.0f);
}
if (inferenceValue.contains("noise_w")) {
synthesisConfig.noiseW = inferenceValue.value("noise_w", 0.8f);
}
if (inferenceValue.contains("phoneme_silence")) {
// phoneme -> seconds of silence to add after
synthesisConfig.phonemeSilenceSeconds.emplace();
auto phonemeSilenceValue = inferenceValue["phoneme_silence"];
for (auto &phonemeItem : phonemeSilenceValue.items()) {
std::string phonemeStr = phonemeItem.key();
if (!isSingleCodepoint(phonemeStr)) {
spdlog::error("\"{}\" is not a single codepoint", phonemeStr);
throw std::runtime_error(
"Phonemes must be one codepoint (phoneme silence)");
}
auto phoneme = getCodepoint(phonemeStr);
(*synthesisConfig.phonemeSilenceSeconds)[phoneme] =
phonemeItem.value().get<float>();
}
} // if phoneme_silence
} // if inference
} /* parseSynthesisConfig */
void parseModelConfig(json &configRoot, ModelConfig &modelConfig) {
modelConfig.numSpeakers = configRoot["num_speakers"].get<SpeakerId>();
if (configRoot.contains("speaker_id_map")) {
if (!modelConfig.speakerIdMap) {
modelConfig.speakerIdMap.emplace();
}
auto speakerIdMapValue = configRoot["speaker_id_map"];
for (auto &speakerItem : speakerIdMapValue.items()) {
std::string speakerName = speakerItem.key();
(*modelConfig.speakerIdMap)[speakerName] =
speakerItem.value().get<SpeakerId>();
}
}
} /* parseModelConfig */
void initialize(PiperConfig &config) {
if (config.useESpeak) {
// Set up espeak-ng for calling espeak_TextToPhonemesWithTerminator
// See: https://github.com/rhasspy/espeak-ng
spdlog::debug("Initializing eSpeak");
int result = espeak_Initialize(AUDIO_OUTPUT_SYNCHRONOUS,
/*buflength*/ 0,
/*path*/ config.eSpeakDataPath.c_str(),
/*options*/ 0);
if (result < 0) {
throw std::runtime_error("Failed to initialize eSpeak-ng");
}
spdlog::debug("Initialized eSpeak");
}
// Load onnx model for libtashkeel
// https://github.com/mush42/libtashkeel/
if (config.useTashkeel) {
spdlog::debug("Using libtashkeel for diacritization");
if (!config.tashkeelModelPath) {
throw std::runtime_error("No path to libtashkeel model");
}
spdlog::debug("Loading libtashkeel model from {}",
config.tashkeelModelPath.value());
config.tashkeelState = std::make_unique<tashkeel::State>();
tashkeel::tashkeel_load(config.tashkeelModelPath.value(),
*config.tashkeelState);
spdlog::debug("Initialized libtashkeel");
}
spdlog::info("Initialized piper");
}
void terminate(PiperConfig &config) {
if (config.useESpeak) {
// Clean up espeak-ng
spdlog::debug("Terminating eSpeak");
espeak_Terminate();
spdlog::debug("Terminated eSpeak");
}
spdlog::info("Terminated piper");
}
void loadModel(std::string modelPath, ModelSession &session, bool useCuda) {
spdlog::debug("Loading onnx model from {}", modelPath);
session.env = Ort::Env(OrtLoggingLevel::ORT_LOGGING_LEVEL_WARNING,
instanceName.c_str());
session.env.DisableTelemetryEvents();
if (useCuda) {
// Use CUDA provider
OrtCUDAProviderOptions cuda_options{};
cuda_options.cudnn_conv_algo_search = OrtCudnnConvAlgoSearchHeuristic;
session.options.AppendExecutionProvider_CUDA(cuda_options);
}
// Slows down performance by ~2x
// session.options.SetIntraOpNumThreads(1);
// Roughly doubles load time for no visible inference benefit
// session.options.SetGraphOptimizationLevel(
// GraphOptimizationLevel::ORT_ENABLE_EXTENDED);
session.options.SetGraphOptimizationLevel(
GraphOptimizationLevel::ORT_DISABLE_ALL);
// Slows down performance very slightly
// session.options.SetExecutionMode(ExecutionMode::ORT_PARALLEL);
session.options.DisableCpuMemArena();
session.options.DisableMemPattern();
session.options.DisableProfiling();
auto startTime = std::chrono::steady_clock::now();
#ifdef _WIN32
auto modelPathW = std::wstring(modelPath.begin(), modelPath.end());
auto modelPathStr = modelPathW.c_str();
#else
auto modelPathStr = modelPath.c_str();
#endif
session.onnx = Ort::Session(session.env, modelPathStr, session.options);
auto endTime = std::chrono::steady_clock::now();
spdlog::debug("Loaded onnx model in {} second(s)",
std::chrono::duration<double>(endTime - startTime).count());
}
// Load Onnx model and JSON config file
void loadVoice(PiperConfig &config, std::string modelPath,
std::string modelConfigPath, Voice &voice,
std::optional<SpeakerId> &speakerId, bool useCuda) {
spdlog::debug("Parsing voice config at {}", modelConfigPath);
std::ifstream modelConfigFile(modelConfigPath);
voice.configRoot = json::parse(modelConfigFile);
parsePhonemizeConfig(voice.configRoot, voice.phonemizeConfig);
parseSynthesisConfig(voice.configRoot, voice.synthesisConfig);
parseModelConfig(voice.configRoot, voice.modelConfig);
if (voice.modelConfig.numSpeakers > 1) {
// Multi-speaker model
if (speakerId) {
voice.synthesisConfig.speakerId = speakerId;
} else {
// Default speaker
voice.synthesisConfig.speakerId = 0;
}
}
spdlog::debug("Voice contains {} speaker(s)", voice.modelConfig.numSpeakers);
loadModel(modelPath, voice.session, useCuda);
} /* loadVoice */
// Phoneme ids to WAV audio
void synthesize(std::vector<PhonemeId> &phonemeIds,
SynthesisConfig &synthesisConfig, ModelSession &session,
std::vector<int16_t> &audioBuffer, SynthesisResult &result) {
spdlog::debug("Synthesizing audio for {} phoneme id(s)", phonemeIds.size());
auto memoryInfo = Ort::MemoryInfo::CreateCpu(
OrtAllocatorType::OrtArenaAllocator, OrtMemType::OrtMemTypeDefault);
// Allocate
std::vector<int64_t> phonemeIdLengths{(int64_t)phonemeIds.size()};
std::vector<float> scales{synthesisConfig.noiseScale,
synthesisConfig.lengthScale,
synthesisConfig.noiseW};
std::vector<Ort::Value> inputTensors;
std::vector<int64_t> phonemeIdsShape{1, (int64_t)phonemeIds.size()};
inputTensors.push_back(Ort::Value::CreateTensor<int64_t>(
memoryInfo, phonemeIds.data(), phonemeIds.size(), phonemeIdsShape.data(),
phonemeIdsShape.size()));
std::vector<int64_t> phomemeIdLengthsShape{(int64_t)phonemeIdLengths.size()};
inputTensors.push_back(Ort::Value::CreateTensor<int64_t>(
memoryInfo, phonemeIdLengths.data(), phonemeIdLengths.size(),
phomemeIdLengthsShape.data(), phomemeIdLengthsShape.size()));
std::vector<int64_t> scalesShape{(int64_t)scales.size()};
inputTensors.push_back(
Ort::Value::CreateTensor<float>(memoryInfo, scales.data(), scales.size(),
scalesShape.data(), scalesShape.size()));
// Add speaker id.
// NOTE: These must be kept outside the "if" below to avoid being deallocated.
std::vector<int64_t> speakerId{
(int64_t)synthesisConfig.speakerId.value_or(0)};
std::vector<int64_t> speakerIdShape{(int64_t)speakerId.size()};
if (synthesisConfig.speakerId) {
inputTensors.push_back(Ort::Value::CreateTensor<int64_t>(
memoryInfo, speakerId.data(), speakerId.size(), speakerIdShape.data(),
speakerIdShape.size()));
}
// From export_onnx.py
std::array<const char *, 4> inputNames = {"input", "input_lengths", "scales",
"sid"};
std::array<const char *, 1> outputNames = {"output"};
// Infer
auto startTime = std::chrono::steady_clock::now();
auto outputTensors = session.onnx.Run(
Ort::RunOptions{nullptr}, inputNames.data(), inputTensors.data(),
inputTensors.size(), outputNames.data(), outputNames.size());
auto endTime = std::chrono::steady_clock::now();
if ((outputTensors.size() != 1) || (!outputTensors.front().IsTensor())) {
throw std::runtime_error("Invalid output tensors");
}
auto inferDuration = std::chrono::duration<double>(endTime - startTime);
result.inferSeconds = inferDuration.count();
const float *audio = outputTensors.front().GetTensorData<float>();
auto audioShape =
outputTensors.front().GetTensorTypeAndShapeInfo().GetShape();
int64_t audioCount = audioShape[audioShape.size() - 1];
result.audioSeconds = (double)audioCount / (double)synthesisConfig.sampleRate;
result.realTimeFactor = 0.0;
if (result.audioSeconds > 0) {
result.realTimeFactor = result.inferSeconds / result.audioSeconds;
}
spdlog::debug("Synthesized {} second(s) of audio in {} second(s)",
result.audioSeconds, result.inferSeconds);
// Get max audio value for scaling
float maxAudioValue = 0.01f;
for (int64_t i = 0; i < audioCount; i++) {
float audioValue = abs(audio[i]);
if (audioValue > maxAudioValue) {
maxAudioValue = audioValue;
}
}
// We know the size up front
audioBuffer.reserve(audioCount);
// Scale audio to fill range and convert to int16
float audioScale = (MAX_WAV_VALUE / std::max(0.01f, maxAudioValue));
for (int64_t i = 0; i < audioCount; i++) {
int16_t intAudioValue = static_cast<int16_t>(
std::clamp(audio[i] * audioScale,
static_cast<float>(std::numeric_limits<int16_t>::min()),
static_cast<float>(std::numeric_limits<int16_t>::max())));
audioBuffer.push_back(intAudioValue);
}
// Clean up
for (std::size_t i = 0; i < outputTensors.size(); i++) {
Ort::detail::OrtRelease(outputTensors[i].release());
}
for (std::size_t i = 0; i < inputTensors.size(); i++) {
Ort::detail::OrtRelease(inputTensors[i].release());
}
}
// ----------------------------------------------------------------------------
// Phonemize text and synthesize audio
void textToAudio(PiperConfig &config, Voice &voice, std::string text,
std::vector<int16_t> &audioBuffer, SynthesisResult &result,
const std::function<void()> &audioCallback) {
std::size_t sentenceSilenceSamples = 0;
if (voice.synthesisConfig.sentenceSilenceSeconds > 0) {
sentenceSilenceSamples = (std::size_t)(
voice.synthesisConfig.sentenceSilenceSeconds *
voice.synthesisConfig.sampleRate * voice.synthesisConfig.channels);
}
if (config.useTashkeel) {
if (!config.tashkeelState) {
throw std::runtime_error("Tashkeel model is not loaded");
}
spdlog::debug("Diacritizing text with libtashkeel: {}", text);
text = tashkeel::tashkeel_run(text, *config.tashkeelState);
}
// Phonemes for each sentence
spdlog::debug("Phonemizing text: {}", text);
std::vector<std::vector<Phoneme>> phonemes;
if (voice.phonemizeConfig.phonemeType == eSpeakPhonemes) {
// Use espeak-ng for phonemization
eSpeakPhonemeConfig eSpeakConfig;
eSpeakConfig.voice = voice.phonemizeConfig.eSpeak.voice;
phonemize_eSpeak(text, eSpeakConfig, phonemes);
} else {
// Use UTF-8 codepoints as "phonemes"
CodepointsPhonemeConfig codepointsConfig;
phonemize_codepoints(text, codepointsConfig, phonemes);
}
// Synthesize each sentence independently.
std::vector<PhonemeId> phonemeIds;
std::map<Phoneme, std::size_t> missingPhonemes;
for (auto phonemesIter = phonemes.begin(); phonemesIter != phonemes.end();
++phonemesIter) {
std::vector<Phoneme> &sentencePhonemes = *phonemesIter;
if (spdlog::should_log(spdlog::level::debug)) {
// DEBUG log for phonemes
std::string phonemesStr;
for (auto phoneme : sentencePhonemes) {
utf8::append(phoneme, std::back_inserter(phonemesStr));
}
spdlog::debug("Converting {} phoneme(s) to ids: {}",
sentencePhonemes.size(), phonemesStr);
}
std::vector<std::shared_ptr<std::vector<Phoneme>>> phrasePhonemes;
std::vector<SynthesisResult> phraseResults;
std::vector<size_t> phraseSilenceSamples;
// Use phoneme/id map from config
PhonemeIdConfig idConfig;
idConfig.phonemeIdMap =
std::make_shared<PhonemeIdMap>(voice.phonemizeConfig.phonemeIdMap);
if (voice.synthesisConfig.phonemeSilenceSeconds) {
// Split into phrases
std::map<Phoneme, float> &phonemeSilenceSeconds =
*voice.synthesisConfig.phonemeSilenceSeconds;
auto currentPhrasePhonemes = std::make_shared<std::vector<Phoneme>>();
phrasePhonemes.push_back(currentPhrasePhonemes);
for (auto sentencePhonemesIter = sentencePhonemes.begin();
sentencePhonemesIter != sentencePhonemes.end();
sentencePhonemesIter++) {
Phoneme ¤tPhoneme = *sentencePhonemesIter;
currentPhrasePhonemes->push_back(currentPhoneme);
if (phonemeSilenceSeconds.count(currentPhoneme) > 0) {
// Split at phrase boundary
phraseSilenceSamples.push_back(
(std::size_t)(phonemeSilenceSeconds[currentPhoneme] *
voice.synthesisConfig.sampleRate *
voice.synthesisConfig.channels));
currentPhrasePhonemes = std::make_shared<std::vector<Phoneme>>();
phrasePhonemes.push_back(currentPhrasePhonemes);
}
}
} else {
// Use all phonemes
phrasePhonemes.push_back(
std::make_shared<std::vector<Phoneme>>(sentencePhonemes));
}
// Ensure results/samples are the same size
while (phraseResults.size() < phrasePhonemes.size()) {
phraseResults.emplace_back();
}
while (phraseSilenceSamples.size() < phrasePhonemes.size()) {
phraseSilenceSamples.push_back(0);
}
// phonemes -> ids -> audio
for (size_t phraseIdx = 0; phraseIdx < phrasePhonemes.size(); phraseIdx++) {
if (phrasePhonemes[phraseIdx]->size() <= 0) {
continue;
}
// phonemes -> ids
phonemes_to_ids(*(phrasePhonemes[phraseIdx]), idConfig, phonemeIds,
missingPhonemes);
if (spdlog::should_log(spdlog::level::debug)) {
// DEBUG log for phoneme ids
std::stringstream phonemeIdsStr;
for (auto phonemeId : phonemeIds) {
phonemeIdsStr << phonemeId << ", ";
}
spdlog::debug("Converted {} phoneme(s) to {} phoneme id(s): {}",
phrasePhonemes[phraseIdx]->size(), phonemeIds.size(),
phonemeIdsStr.str());
}
// ids -> audio
synthesize(phonemeIds, voice.synthesisConfig, voice.session, audioBuffer,
phraseResults[phraseIdx]);
// Add end of phrase silence
for (std::size_t i = 0; i < phraseSilenceSamples[phraseIdx]; i++) {
audioBuffer.push_back(0);
}
result.audioSeconds += phraseResults[phraseIdx].audioSeconds;
result.inferSeconds += phraseResults[phraseIdx].inferSeconds;
phonemeIds.clear();
}
// Add end of sentence silence
if (sentenceSilenceSamples > 0) {
for (std::size_t i = 0; i < sentenceSilenceSamples; i++) {
audioBuffer.push_back(0);
}
}
if (audioCallback) {
// Call back must copy audio since it is cleared afterwards.
audioCallback();
audioBuffer.clear();
}
phonemeIds.clear();
}
if (missingPhonemes.size() > 0) {
spdlog::warn("Missing {} phoneme(s) from phoneme/id map!",
missingPhonemes.size());
for (auto phonemeCount : missingPhonemes) {
std::string phonemeStr;
utf8::append(phonemeCount.first, std::back_inserter(phonemeStr));
spdlog::warn("Missing \"{}\" (\\u{:04X}): {} time(s)", phonemeStr,
(uint32_t)phonemeCount.first, phonemeCount.second);
}
}
if (result.audioSeconds > 0) {
result.realTimeFactor = result.inferSeconds / result.audioSeconds;
}
} /* textToAudio */
// Phonemize text and synthesize audio to WAV file
void textToWavFile(PiperConfig &config, Voice &voice, std::string text,
std::ostream &audioFile, SynthesisResult &result) {
std::vector<int16_t> audioBuffer;
textToAudio(config, voice, text, audioBuffer, result, NULL);
// Write WAV
auto synthesisConfig = voice.synthesisConfig;
writeWavHeader(synthesisConfig.sampleRate, synthesisConfig.sampleWidth,
synthesisConfig.channels, (int32_t)audioBuffer.size(),
audioFile);
audioFile.write((const char *)audioBuffer.data(),
sizeof(int16_t) * audioBuffer.size());
} /* textToWavFile */
} // namespace piper