49 lines
1.3 KiB
C++
49 lines
1.3 KiB
C++
// sherpa-onnx/csrc/offline-tts-vits-model-config.h
|
|
//
|
|
// Copyright (c) 2023 Xiaomi Corporation
|
|
|
|
#ifndef SHERPA_ONNX_CSRC_OFFLINE_TTS_VITS_MODEL_CONFIG_H_
|
|
#define SHERPA_ONNX_CSRC_OFFLINE_TTS_VITS_MODEL_CONFIG_H_
|
|
|
|
#include <string>
|
|
|
|
#include "sherpa-onnx/csrc/parse-options.h"
|
|
|
|
namespace sherpa_onnx {
|
|
|
|
struct OfflineTtsVitsModelConfig {
|
|
std::string model;
|
|
std::string lexicon;
|
|
std::string tokens;
|
|
|
|
float noise_scale = 0.667;
|
|
float noise_scale_w = 0.8;
|
|
float length_scale = 1;
|
|
|
|
// used only for multi-speaker models, e.g, vctk speech dataset.
|
|
// Not applicable for single-speaker models, e.g., ljspeech dataset
|
|
|
|
OfflineTtsVitsModelConfig() = default;
|
|
|
|
OfflineTtsVitsModelConfig(const std::string &model,
|
|
const std::string &lexicon,
|
|
const std::string &tokens,
|
|
float noise_scale = 0.667,
|
|
float noise_scale_w = 0.8, float length_scale = 1)
|
|
: model(model),
|
|
lexicon(lexicon),
|
|
tokens(tokens),
|
|
noise_scale(noise_scale),
|
|
noise_scale_w(noise_scale_w),
|
|
length_scale(length_scale) {}
|
|
|
|
void Register(ParseOptions *po);
|
|
bool Validate() const;
|
|
|
|
std::string ToString() const;
|
|
};
|
|
|
|
} // namespace sherpa_onnx
|
|
|
|
#endif // SHERPA_ONNX_CSRC_OFFLINE_TTS_VITS_MODEL_CONFIG_H_
|