This model was trained by kan-bayashi using jsut/tts1 recipe in espnet. Python API See https://github.com/espnet/espnet_model_zoo Evaluate in the recipe git clone https://github.com/espnet/espnet cd espnet git checkout a15fd3b33492e67d06e39de5c05f4a126af75518 pip install -e . cd egs2/jsut/tts1 # Download the model file here ./run.sh --skip_data_prep false --skip_train true --download_model kan-bayashi/jsut_tts_train_fastspeech2_raw_phn_jaconv_pyopenjtalk_train.loss.best Config config: conf/tuning/train_fastspeech2.yaml print_config: false log_level: INFO dry_run: false iterator_type: sequence output_dir: exp/tts_train_fastspeech2_raw_phn_jaconv_pyopenjtalk ngpu: 1 seed: 0 num_workers: 1 num_att_plot: 3 dist_backend: nccl dist_init_method: env:// dist_world_size: null dist_rank: null local_rank: 0 dist_master_addr: null dist_master_port: null dist_launcher: null multiprocessing_distributed: false cudnn_enabled: true cudnn_benchmark: false cudnn_deterministic: true collect_stats: false write_collected_feats: false max_epoch: 1000 patience: null val_scheduler_criterion: - valid - loss early_stopping_criterion: - valid - loss - min best_model_criterion: - - valid - loss - min - - train - loss - min keep_nbest_models: 5 grad_clip: 1.0 grad_noise: false accum_grad: 8 no_forward_run: false resume: true train_dtype: float32 log_interval: null pretrain_path: [] pretrain_key: [] num_iters_per_epoch: null batch_size: 20 valid_batch_size: null batch_bins: 3000000 valid_batch_bins: null train_shape_file: - exp/tts_stats_raw_phn_jaconv_pyopenjtalk/train/text_shape.phn - exp/tts_stats_raw_phn_jaconv_pyopenjtalk/train/speech_shape valid_shape_file: - exp/tts_stats_raw_phn_jaconv_pyopenjtalk/valid/text_shape.phn - exp/tts_stats_raw_phn_jaconv_pyopenjtalk/valid/speech_shape batch_type: numel valid_batch_type: null fold_length: - 150 - 240000 sort_in_batch: descending sort_batch: descending multiple_iterator: false chunk_length: 500 chunk_shift_ratio: 0.5 num_cache_chunks: 1024 train_data_path_and_name_and_type: - - dump/raw/tr_no_dev/text - text - text - - exp/tts_train_tacotron2_raw_phn_jaconv_pyopenjtalk/decode_tacotron2_teacher_forcing_train.loss.best/tr_no_dev/durations - durations - text_int - - dump/raw/tr_no_dev/wav.scp - speech - sound - - exp/tts_stats_raw_phn_jaconv_pyopenjtalk/train/collect_feats/pitch.scp - pitch - npy - - exp/tts_stats_raw_phn_jaconv_pyopenjtalk/train/collect_feats/energy.scp - energy - npy valid_data_path_and_name_and_type: - - dump/raw/dev/text - text - text - - exp/tts_train_tacotron2_raw_phn_jaconv_pyopenjtalk/decode_tacotron2_teacher_forcing_train.loss.best/dev/durations - durations - text_int - - dump/raw/dev/wav.scp - speech - sound - - exp/tts_stats_raw_phn_jaconv_pyopenjtalk/valid/collect_feats/pitch.scp - pitch - npy - - exp/tts_stats_raw_phn_jaconv_pyopenjtalk/valid/collect_feats/energy.scp - energy - npy allow_variable_data_keys: false max_cache_size: 0.0 valid_max_cache_size: null optim: adam optim_conf: lr: 1.0 scheduler: noamlr scheduler_conf: model_size: 384 warmup_steps: 4000 token_list: - - - ty - dy - v - py - my - by - ny - hy - gy - ry - ky - f - p - z - ch - ts - j - b - y - h - cl - I - U - w - g - d - sh - pau - m - N - s - r - t - n - k - e - u - i - o - a - odim: null model_conf: {} use_preprocessor: true token_type: phn bpemodel: null non_linguistic_symbols: null cleaner: jaconv g2p: pyopenjtalk feats_extract: fbank feats_extract_conf: fs: 24000 fmin: 80 fmax: 7600 n_mels: 80 hop_length: 300 n_fft: 2048 win_length: 1200 normalize: global_mvn normalize_conf: stats_file: exp/tts_stats_raw_phn_jaconv_pyopenjtalk/train/feats_stats.npz tts: fastspeech2 tts_conf: adim: 384 aheads: 2 elayers: 4 eunits: 1536 dlayers: 4 dunits: 1536 positionwise_layer_type: conv1d positionwise_conv_kernel_size: 3 duration_predictor_layers: 2 duration_predictor_chans: 256 duration_predictor_kernel_size: 3 postnet_layers: 5 postnet_filts: 5 postnet_chans: 256 use_masking: true use_scaled_pos_enc: true encoder_normalize_before: false decoder_normalize_before: false reduction_factor: 1 init_type: xavier_uniform init_enc_alpha: 1.0 init_dec_alpha: 1.0 transformer_enc_dropout_rate: 0.2 transformer_enc_positional_dropout_rate: 0.2 transformer_enc_attn_dropout_rate: 0.2 transformer_dec_dropout_rate: 0.2 transformer_dec_positional_dropout_rate: 0.2 transformer_dec_attn_dropout_rate: 0.2 pitch_predictor_layers: 2 pitch_predictor_chans: 256 pitch_predictor_kernel_size: 3 pitch_predictor_dropout: 0.5 pitch_embed_kernel_size: 1 pitch_embed_dropout: 0.0 stop_gradient_from_pitch_predictor: false energy_predictor_layers: 5 energy_predictor_chans: 256 energy_predictor_kernel_size: 5 energy_predictor_dropout: 0.5 energy_embed_kernel_size: 1 energy_embed_dropout: 0.0 stop_gradient_from_energy_predictor: true pitch_extract: dio pitch_extract_conf: fs: 24000 n_fft: 2048 hop_length: 300 f0max: 800 f0min: 80 pitch_normalize: global_mvn pitch_normalize_conf: stats_file: exp/tts_stats_raw_phn_jaconv_pyopenjtalk/train/pitch_stats.npz energy_extract: energy energy_extract_conf: fs: 24000 n_fft: 2048 hop_length: 300 win_length: 1200 energy_normalize: global_mvn energy_normalize_conf: stats_file: exp/tts_stats_raw_phn_jaconv_pyopenjtalk/train/energy_stats.npz required: - output_dir - token_list distributed: false