config/vitssvc.json

{
 "base_config": "config/svc/base.json",
 "model_type": "VITS",
 "task_type": "svc",
 "preprocess": {
 // Config for features extraction
 "extract_mel": true,
 "extract_pitch": true,
 "pitch_extractor": "parselmouth",
 "extract_energy": true,
 "extract_uv": true,
 "extract_linear_spec": true,
 "extract_audio": true,
 "mel_min_max_norm": true,
 // Config for features usage
 "use_linear": true,
 "use_mel": true,
 "use_min_max_norm_mel": false,
 "use_audio": true,
 "use_frame_pitch": true,
 "use_uv": true,
 "use_spkid": true,
 "use_contentvec": false,
 "use_whisper": false,
 "use_wenet": false,
 "use_text": false,
 "use_phone": false,
 "fmin": 0,
 "fmax": 12000,
 "f0_min": 50,
 "f0_max": 1100,
 // f0_bin in sovits
 "pitch_bin": 256,
 // filter_length in sovits
 "n_fft": 1024,
 // hop_length in sovits
 "hop_size": 256,
 // win_length in sovits
 "win_size": 1024,
 "segment_size": 8192,
 "n_mel": 100,
 "sample_rate": 24000,
 "mel_min_max_stats_dir": "mel_min_max_stats",
 "whisper_dir": "whisper",
 "contentvec_dir": "contentvec",
 "wenet_dir": "wenet",
 "mert_dir": "mert",
 // Meta file
 "train_file": "train.json",
 "valid_file": "test.json",
 "spk2id": "singers.json",
 "utt2spk": "utt2singer"
 },
 "model": {
 "condition_encoder": {
 "merge_mode": "add",
 "input_melody_dim": 1,
 "use_log_f0": true,
 "n_bins_melody": 256,
 "output_melody_dim": 384,
 "input_loudness_dim": 1,
 "use_log_loudness": true,
 "n_bins_loudness": 256,
 "output_loudness_dim": 384,
 "use_whisper": false,
 "use_contentvec": false,
 "use_wenet": false,
 "use_mert": false,
 "whisper_dim": 1024,
 "contentvec_dim": 256,
 "mert_dim": 256,
 "wenet_dim": 512,
 "content_encoder_dim": 384,
 "singer_table_size": 512,
 "output_singer_dim": 384,
 "output_content_dim": 384,
 "use_spkid": true,
 "pitch_max": 1100.0,
 "pitch_min": 50.0,
 },
 "vits": {
 "filter_channels": 256,
 "gin_channels": 256,
 "hidden_channels": 384,
 "inter_channels": 384,
 "kernel_size": 3,
 "n_flow_layer": 4,
 "n_heads": 2,
 "n_layers": 6,
 "n_layers_q": 3,
 "n_speakers": 512,
 "p_dropout": 0.1,
 "use_spectral_norm": false,
 },
 "generator": "hifigan",
 "generator_config": {
 "hifigan": {
 "resblock": "1",
 "resblock_kernel_sizes": [
 3,
 7,
 11
 ],
 "upsample_rates": [
 8,
 8,
 2,
 2
 ],
 "upsample_kernel_sizes": [
 16,
 16,
 4,
 4
 ],
 "upsample_initial_channel": 512,
 "resblock_dilation_sizes": [
 [
 1,
 3,
 5
 ],
 [
 1,
 3,
 5
 ],
 [
 1,
 3,
 5
 ]
 ]
 },
 "melgan": {
 "ratios": [
 8,
 8,
 2,
 2
 ],
 "ngf": 32,
 "n_residual_layers": 3,
 "num_D": 3,
 "ndf": 16,
 "n_layers": 4,
 "downsampling_factor": 4
 },
 "bigvgan": {
 "resblock": "1",
 "activation": "snakebeta",
 "snake_logscale": true,
 "upsample_rates": [
 8,
 8,
 2,
 2
 ],
 "upsample_kernel_sizes": [
 16,
 16,
 4,
 4
 ],
 "upsample_initial_channel": 512,
 "resblock_kernel_sizes": [
 3,
 7,
 11
 ],
 "resblock_dilation_sizes": [
 [
 1,
 3,
 5
 ],
 [
 1,
 3,
 5
 ],
 [
 1,
 3,
 5
 ]
 ]
 },
 "nsfhifigan": {
 "resblock": "1",
 "harmonic_num": 8,
 "upsample_rates": [
 8,
 8,
 2,
 2
 ],
 "upsample_kernel_sizes": [
 16,
 16,
 4,
 4
 ],
 "upsample_initial_channel": 768,
 "resblock_kernel_sizes": [
 3,
 7,
 11
 ],
 "resblock_dilation_sizes": [
 [
 1,
 3,
 5
 ],
 [
 1,
 3,
 5
 ],
 [
 1,
 3,
 5
 ]
 ]
 },
 "apnet": {
 "ASP_channel": 512,
 "ASP_resblock_kernel_sizes": [
 3,
 7,
 11
 ],
 "ASP_resblock_dilation_sizes": [
 [
 1,
 3,
 5
 ],
 [
 1,
 3,
 5
 ],
 [
 1,
 3,
 5
 ]
 ],
 "ASP_input_conv_kernel_size": 7,
 "ASP_output_conv_kernel_size": 7,
 "PSP_channel": 512,
 "PSP_resblock_kernel_sizes": [
 3,
 7,
 11
 ],
 "PSP_resblock_dilation_sizes": [
 [
 1,
 3,
 5
 ],
 [
 1,
 3,
 5
 ],
 [
 1,
 3,
 5
 ]
 ],
 "PSP_input_conv_kernel_size": 7,
 "PSP_output_R_conv_kernel_size": 7,
 "PSP_output_I_conv_kernel_size": 7,
 }
 },
 },
 "train": {
 "fp16_run": true,
 "learning_rate": 2e-4,
 "betas": [
 0.8,
 0.99
 ],
 "eps": 1e-9,
 "batch_size": 16,
 "lr_decay": 0.999875,
 // "segment_size": 8192,
 "init_lr_ratio": 1,
 "warmup_epochs": 0,
 "c_mel": 45,
 "c_kl": 1.0,
 "AdamW": {
 "betas": [
 0.8,
 0.99
 ],
 "eps": 1e-9,
 }
 }
}