GPT-SoVITS模型|《红楼梦》林黛玉声音配音模型

人物介绍

林黛玉是中国古典名著《红楼梦》中的女主角之一，由清代作家曹雪芹所创作。她以其才情出众、性格敏感多愁、体弱多病而著称，是小说中一个极具深度和复杂性的角色。关于林黛玉的声音特点，87版《红楼梦》中，林黛玉由张海玲配音，其声音特点可概括为：温婉细腻，与角色温柔多情的性格相符；略带哀愁，流露出对世态炎凉、命运无常的感慨；清晰悦耳，展现深厚文化底蕴；情感丰富，能准确传达各种情绪，深刻反映内心世界。

温婉细腻：林黛玉作为一位才情出众、性格敏感多愁的女性角色，其声音很可能带有一种温婉细腻的感觉，与她温柔多情的性格相契合。
略带哀愁：由于林黛玉性格中带有深深的哀愁与忧郁，她的声音在表达情感时可能会不自觉地流露出这种情绪。这种哀愁并非简单的悲伤，而是融合了对世态炎凉的感慨、对命运无常的无奈以及对美好事物逝去的惋惜。
清晰悦耳：作为一位受过良好教育的大家闺秀，林黛玉的发音应该非常清晰，语调抑扬顿挫，给人以悦耳之感。她的声音中可能还蕴含着一种书卷气，体现出她深厚的文化底蕴和文学修养。

GPT-SoVITS模型配音效果

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

2025-07-06 11:14:20,878  lindaiyu  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': 20, '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': 10, 'gpu_numbers': '0'}, 'data': {'max_wav_value': 32768.0, 'sampling_rate': 32000, 'filter_lenlindaiyuh': 2048, 'hop_lenlindaiyuh': 640, 'win_lenlindaiyuh': 2048, 'n_mel_channels': 128, 'mel_fmin': 0.0, 'mel_fmax': None, 'add_blank': True, 'n_speakers': 300, 'cleaned_text': True, 'exp_dir': 'logs/lindaiyu'}, '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/lindaiyu', 'content_module': 'cnhubert', 'save_weight_dir': 'SoVITS_weights', 'name': 'lindaiyu', 'pretrain': None, 'resume_step': None}
2025-07-06 11:14:22,336  lindaiyu  INFO  loaded pretrained GPT_SoVITS/pretrained_models/s2G488k.pth
2025-07-06 11:14:22,539  lindaiyu  INFO  loaded pretrained GPT_SoVITS/pretrained_models/s2D488k.pth
2025-07-06 11:14:54,947  lindaiyu  INFO  Train Epoch: 1 [0%]
2025-07-06 11:14:54,948  lindaiyu  INFO  [2.5188231468200684, 2.587672233581543, 7.5846028327941895, 21.921340942382812, 0.0, 2.863032102584839, 0, 9.99875e-05]
2025-07-06 11:15:08,761  lindaiyu  INFO  ====> Epoch: 1
2025-07-06 11:15:28,200  lindaiyu  INFO  ====> Epoch: 2
2025-07-06 11:15:44,386  lindaiyu  INFO  ====> Epoch: 3
2025-07-06 11:16:00,712  lindaiyu  INFO  ====> Epoch: 4
2025-07-06 11:16:17,322  lindaiyu  INFO  ====> Epoch: 5
2025-07-06 11:16:31,917  lindaiyu  INFO  ====> Epoch: 6
2025-07-06 11:16:47,588  lindaiyu  INFO  ====> Epoch: 7
2025-07-06 11:17:03,477  lindaiyu  INFO  ====> Epoch: 8
2025-07-06 11:17:20,315  lindaiyu  INFO  ====> Epoch: 9
2025-07-06 11:17:35,272  lindaiyu  INFO  Saving model and optimizer state at iteration 10 to logs/lindaiyu/logs_s2\G_233333333333.pth
2025-07-06 11:17:36,166  lindaiyu  INFO  Saving model and optimizer state at iteration 10 to logs/lindaiyu/logs_s2\D_233333333333.pth
2025-07-06 11:17:38,124  lindaiyu  INFO  saving ckpt lindaiyu_e10:Success.
2025-07-06 11:17:38,124  lindaiyu  INFO  ====> Epoch: 10
2025-07-06 11:17:54,918  lindaiyu  INFO  ====> Epoch: 11
2025-07-06 11:18:01,914  lindaiyu  INFO  Train Epoch: 12 [11%]
2025-07-06 11:18:01,915  lindaiyu  INFO  [2.7913379669189453, 2.175286293029785, 8.353038787841797, 20.010730743408203, 0.0, 2.0773844718933105, 100, 9.98501030820433e-05]
2025-07-06 11:18:13,304  lindaiyu  INFO  ====> Epoch: 12
2025-07-06 11:18:31,755  lindaiyu  INFO  ====> Epoch: 13
2025-07-06 11:18:50,774  lindaiyu  INFO  ====> Epoch: 14
2025-07-06 11:19:05,630  lindaiyu  INFO  ====> Epoch: 15
2025-07-06 11:19:24,356  lindaiyu  INFO  ====> Epoch: 16
2025-07-06 11:19:39,803  lindaiyu  INFO  ====> Epoch: 17
2025-07-06 11:19:53,730  lindaiyu  INFO  ====> Epoch: 18
2025-07-06 11:20:09,730  lindaiyu  INFO  ====> Epoch: 19
2025-07-06 11:20:25,700  lindaiyu  INFO  Saving model and optimizer state at iteration 20 to logs/lindaiyu/logs_s2\G_233333333333.pth
2025-07-06 11:20:26,871  lindaiyu  INFO  Saving model and optimizer state at iteration 20 to logs/lindaiyu/logs_s2\D_233333333333.pth
2025-07-06 11:20:30,170  lindaiyu  INFO  saving ckpt lindaiyu_e20:Success.
2025-07-06 11:20:30,170  lindaiyu  INFO  ====> Epoch: 20