GPT-SOVITS模型|小笛声音模型小笛配音模型

10天前更新

100

擅长领域：童声/情感/影视/动漫可爱男孩，自然动听

配音效果

鉴于GPT-SOVITS模型自回归特性，即其配音情绪高度依赖于所提供的参考音频，特此说明：本视频所展示的配音情绪仅为采用某一特定参考音频时的效果示例，并不全面反映GPT-SOVITS模型能够生成的全部情绪范围及最终配音质量的上限。模型的最终表现将随着不同参考音频的输入而展现出多样化。

训练日志

2024-07-31 10:12:10,168	魔小笛	INFO	{'train': {'log_interval': 100, 'eval_interval': 500, 'seed': 1234, 'epochs': 20, 'learning_rate': 0.0001, 'betas': [0.8, 0.99], 'eps': 1e-09, 'batch_size': 11, 'fp16_run': True, 'lr_decay': 0.999875, 'segment_size': 20480, 'init_lr_ratio': 1, 'warmup_epochs': 0, 'c_mel': 45, 'c_kl': 1.0, 'text_low_lr_rate': 0.4, 'pretrained_s2G': 'GPT_SoVITS/pretrained_models/s2G488k.pth', 'pretrained_s2D': 'GPT_SoVITS/pretrained_models/s2D488k.pth', 'if_save_latest': True, 'if_save_every_weights': True, 'save_every_epoch': 20, 'gpu_numbers': '0'}, 'data': {'max_wav_value': 32768.0, 'sampling_rate': 32000, 'filter_length': 2048, 'hop_length': 640, 'win_length': 2048, 'n_mel_channels': 128, 'mel_fmin': 0.0, 'mel_fmax': None, 'add_blank': True, 'n_speakers': 300, 'cleaned_text': True, 'exp_dir': 'logs/魔小笛'}, 'model': {'inter_channels': 192, 'hidden_channels': 192, 'filter_channels': 768, 'n_heads': 2, 'n_layers': 6, 'kernel_size': 3, 'p_dropout': 0.1, 'resblock': '1', 'resblock_kernel_sizes': [3, 7, 11], 'resblock_dilation_sizes': [[1, 3, 5], [1, 3, 5], [1, 3, 5]], 'upsample_rates': [10, 8, 2, 2, 2], 'upsample_initial_channel': 512, 'upsample_kernel_sizes': [16, 16, 8, 2, 2], 'n_layers_q': 3, 'use_spectral_norm': False, 'gin_channels': 512, 'semantic_frame_rate': '25hz', 'freeze_quantizer': True}, 's2_ckpt_dir': 'logs/魔小笛', 'content_module': 'cnhubert', 'save_weight_dir': 'SoVITS_weights', 'name': '魔小笛', 'pretrain': None, 'resume_step': None}
2024-07-31 10:12:11,456	魔小笛	INFO	loaded pretrained GPT_SoVITS/pretrained_models/s2G488k.pth
2024-07-31 10:12:11,651	魔小笛	INFO	loaded pretrained GPT_SoVITS/pretrained_models/s2D488k.pth
2024-07-31 10:12:38,037	魔小笛	INFO	Train Epoch: 1 [0%]
2024-07-31 10:12:38,038	魔小笛	INFO	[2.4836320877075195, 2.5544421672821045, 9.588512420654297, 20.640321731567383, 0.0, 2.021796941757202, 0, 9.99875e-05]
2024-07-31 10:12:50,766	魔小笛	INFO	====> Epoch: 1
2024-07-31 10:13:03,631	魔小笛	INFO	====> Epoch: 2
2024-07-31 10:13:16,586	魔小笛	INFO	====> Epoch: 3
2024-07-31 10:13:29,808	魔小笛	INFO	====> Epoch: 4
2024-07-31 10:13:42,433	魔小笛	INFO	====> Epoch: 5
2024-07-31 10:13:55,647	魔小笛	INFO	====> Epoch: 6
2024-07-31 10:14:05,874	魔小笛	INFO	Train Epoch: 7 [67%]
2024-07-31 10:14:05,874	魔小笛	INFO	[2.4023337364196777, 2.4813432693481445, 9.144084930419922, 19.008525848388672, 0.0, 1.8600295782089233, 100, 9.991253280566489e-05]
2024-07-31 10:14:09,028	魔小笛	INFO	====> Epoch: 7
2024-07-31 10:14:22,011	魔小笛	INFO	====> Epoch: 8
2024-07-31 10:14:34,989	魔小笛	INFO	====> Epoch: 9
2024-07-31 10:14:48,868	魔小笛	INFO	====> Epoch: 10
2024-07-31 10:15:01,781	魔小笛	INFO	====> Epoch: 11
2024-07-31 10:15:15,649	魔小笛	INFO	====> Epoch: 12
2024-07-31 10:15:29,586	魔小笛	INFO	====> Epoch: 13
2024-07-31 10:15:36,004	魔小笛	INFO	Train Epoch: 14 [33%]
2024-07-31 10:15:36,004	魔小笛	INFO	[2.2538585662841797, 2.821960687637329, 10.023470878601074, 18.184921264648438, 0.0, 1.2744669914245605, 200, 9.982514211643064e-05]
2024-07-31 10:15:43,018	魔小笛	INFO	====> Epoch: 14
2024-07-31 10:15:56,695	魔小笛	INFO	====> Epoch: 15
2024-07-31 10:16:09,049	魔小笛	INFO	====> Epoch: 16
2024-07-31 10:16:21,862	魔小笛	INFO	====> Epoch: 17
2024-07-31 10:16:34,777	魔小笛	INFO	====> Epoch: 18
2024-07-31 10:16:47,896	魔小笛	INFO	====> Epoch: 19
2024-07-31 10:17:01,016	魔小笛	INFO	Saving model and optimizer state at iteration 20 to logs/魔小笛/logs_s2\G_233333333333.pth
2024-07-31 10:17:01,886	魔小笛	INFO	Saving model and optimizer state at iteration 20 to logs/魔小笛/logs_s2\D_233333333333.pth
2024-07-31 10:17:03,535	魔小笛	INFO	saving ckpt 魔小笛_e20:Success.
2024-07-31 10:17:03,536	魔小笛	INFO	====> Epoch: 20