visinger2-zh-jp-multisinger-svs

This model is espnet/visinger2-zh-jp-multisinger-svs, a multilingual (Chinese and Japanese) singing voice synthesis (SVS) component. It is a key part of the SingingSDS: A Singing-Capable Spoken Dialogue System for Conversational Roleplay Applications project.

• 📚 Paper: SingingSDS: A Singing-Capable Spoken Dialogue System
• 💻 Project Code on GitHub: SingingSDS
• 🚀 Live Demo on Hugging Face Spaces

⚠️ Notice

This model is for research use only and is strictly prohibited for any commercial use. Please make sure to comply with this restriction before using the model.

ESPnet2 SVS model

Demo: How to use in ESPnet2

Follow the ESPnet installation instructions if you haven't done that already.

cd espnet
git checkout c9a3878a76e1b738ccfc3aa96912d14e0e2d3134
pip install -e .
cd egs2/mixed/svs1
./run.sh --skip_data_prep false --skip_train true --download_model espnet/visinger2-zh-jp-multisinger-svs

SVS config

config: conf/tuning/train_visinger2_spk_embed_lang.yaml
print_config: false
log_level: INFO
drop_last_iter: false
dry_run: false
iterator_type: sequence
valid_iterator_type: null
output_dir: exp/svs_train_visinger2_spk_embed_lang_1019
ngpu: 1
seed: 777
num_workers: 0
num_att_plot: 0
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
unused_parameters: true
sharded_ddp: false
use_deepspeed: false
deepspeed_config: null
cudnn_enabled: true
cudnn_benchmark: false
cudnn_deterministic: false
use_tf32: false
collect_stats: false
write_collected_feats: false
max_epoch: 500
patience: null
val_scheduler_criterion:
- valid
- loss
early_stopping_criterion:
- valid
- loss
- min
best_model_criterion:
-   - train
    - total_count
    - max
keep_nbest_models: 15
nbest_averaging_interval: 0
grad_clip: -1
grad_clip_type: 2.0
grad_noise: false
accum_grad: 1
no_forward_run: false
resume: true
train_dtype: float32
use_amp: false
log_interval: 50
use_matplotlib: true
use_tensorboard: true
create_graph_in_tensorboard: false
use_wandb: false
wandb_project: null
wandb_id: null
wandb_entity: null
wandb_name: null
wandb_model_log_interval: -1
detect_anomaly: false
use_adapter: false
adapter: lora
save_strategy: all
adapter_conf: {}
pretrain_path: null
init_param: []
ignore_init_mismatch: false
freeze_param: []
num_iters_per_epoch: 1000
batch_size: 8
valid_batch_size: null
batch_bins: 1000000
valid_batch_bins: null
category_sample_size: 10
train_shape_file:
- exp/svs_stats_raw_phn_None_zh_jp_44100Hz_1019/train/text_shape.phn
- exp/svs_stats_raw_phn_None_zh_jp_44100Hz_1019/train/singing_shape
valid_shape_file:
- exp/svs_stats_raw_phn_None_zh_jp_44100Hz_1019/valid/text_shape.phn
- exp/svs_stats_raw_phn_None_zh_jp_44100Hz_1019/valid/singing_shape
batch_type: sorted
valid_batch_type: null
fold_length:
- 150
- 409600
sort_in_batch: descending
shuffle_within_batch: false
sort_batch: descending
multiple_iterator: false
chunk_length: 500
chunk_shift_ratio: 0.5
num_cache_chunks: 1024
chunk_excluded_key_prefixes: []
chunk_default_fs: null
chunk_max_abs_length: null
chunk_discard_short_samples: true
train_data_path_and_name_and_type:
-   - dump44100_1019/raw/tr_no_dev/text
    - text
    - text
-   - dump44100_1019/raw/tr_no_dev/wav.scp
    - singing
    - sound
-   - dump44100_1019/raw/tr_no_dev/label
    - label
    - duration
-   - dump44100_1019/raw/tr_no_dev/score.scp
    - score
    - score
-   - exp/svs_stats_raw_phn_None_zh_jp_44100Hz_1019/train/collect_feats/pitch.scp
    - pitch
    - npy
-   - exp/svs_stats_raw_phn_None_zh_jp_44100Hz_1019/train/collect_feats/feats.scp
    - feats
    - npy
-   - dump44100_1019/espnet_spk/tr_no_dev/espnet_spk.scp
    - spembs
    - kaldi_ark
-   - dump44100_1019/raw/tr_no_dev/utt2lid
    - lids
    - text_int
valid_data_path_and_name_and_type:
-   - dump44100_1019/raw/dev/text
    - text
    - text
-   - dump44100_1019/raw/dev/wav.scp
    - singing
    - sound
-   - dump44100_1019/raw/dev/label
    - label
    - duration
-   - dump44100_1019/raw/dev/score.scp
    - score
    - score
-   - exp/svs_stats_raw_phn_None_zh_jp_44100Hz_1019/valid/collect_feats/pitch.scp
    - pitch
    - npy
-   - exp/svs_stats_raw_phn_None_zh_jp_44100Hz_1019/valid/collect_feats/feats.scp
    - feats
    - npy
-   - dump44100_1019/espnet_spk/dev/espnet_spk.scp
    - spembs
    - kaldi_ark
-   - dump44100_1019/raw/dev/utt2lid
    - lids
    - text_int
multi_task_dataset: false
allow_variable_data_keys: false
max_cache_size: 0.0
max_cache_fd: 32
allow_multi_rates: false
valid_max_cache_size: null
exclude_weight_decay: false
exclude_weight_decay_conf: {}
optim: adamw
optim_conf:
    lr: 0.0002
    betas:
    - 0.8
    - 0.99
    eps: 1.0e-09
    weight_decay: 0.0
scheduler: exponentiallr
scheduler_conf:
    gamma: 0.998
optim2: adamw
optim2_conf:
    lr: 0.0002
    betas:
    - 0.8
    - 0.99
    eps: 1.0e-09
    weight_decay: 0.0
scheduler2: exponentiallr
scheduler2_conf:
    gamma: 0.998
generator_first: true
skip_discriminator_prob: 0.0
input_size: null
token_list:
-
-
- SP
- i@zh
- e@zh
- d@zh
- y@zh
- uo@zh
- sh@zh
- ian@zh
- ai@zh
- w@zh
- n@zh
- u@zh
- x@zh
- j@zh
- l@zh
- h@zh
- b@zh
- iii@zh
- zh@zh
- uei@zh
- m@zh
- ing@zh
- q@zh
- g@zh
- eng@zh
- z@zh
- a@zh
- an@zh
- en@zh
- iou@zh
- ao@zh
- t@zh
- ou@zh
- iang@zh
- AP
- ong@zh
- ang@zh
- ei@zh
- iao@zh
- ie@zh
- f@zh
- r@zh
- k@zh
- ch@zh
- v@zh
- in@zh
- uan@zh
- c@zh
- uang@zh
- s@zh
- a@jp
- ii@zh
- van@zh
- p@zh
- ve@zh
- o@jp
- i@jp
- ia@zh
- uen@zh
- ua@zh
- iong@zh
- u@jp
- e@jp
- k@jp
- uai@zh
- n@jp
- t@jp
- r@jp
- m@jp
- er@zh
- vn@zh
- s@jp
- d@jp
- w@jp
- o@zh
- sh@jp
- g@jp
- N@jp
- y@jp
- b@jp
- ts@jp
- h@jp
- z@jp