GPT-SoVITS模型 azure 晓晨声音模型

GPT-SoVITS模型微软 azure 晓晨声音模型

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。

模型下载

模型训练日志

2025-07-24 12:17:55,600	xiaochen	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/xiaochen'}, '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/xiaochen', 'content_module': 'cnhubert', 'save_weight_dir': 'SoVITS_weights', 'name': 'xiaochen', 'pretrain': None, 'resume_step': None}
2025-07-24 12:17:57,127	xiaochen	INFO	loaded pretrained GPT_SoVITS/pretrained_models/s2G488k.pth
2025-07-24 12:17:57,336	xiaochen	INFO	loaded pretrained GPT_SoVITS/pretrained_models/s2D488k.pth
2025-07-24 12:18:29,312	xiaochen	INFO	Train Epoch: 1 [0%]
2025-07-24 12:18:29,313	xiaochen	INFO	[2.8378677368164062, 1.877305030822754, 11.843710899353027, 18.460004806518555, 0.0, 1.7773140668869019, 0, 9.99875e-05]
2025-07-24 12:18:44,433	xiaochen	INFO	====> Epoch: 1
2025-07-24 12:18:58,345	xiaochen	INFO	====> Epoch: 2
2025-07-24 12:19:11,447	xiaochen	INFO	====> Epoch: 3
2025-07-24 12:19:24,624	xiaochen	INFO	====> Epoch: 4
2025-07-24 12:19:37,623	xiaochen	INFO	====> Epoch: 5
2025-07-24 12:19:42,863	xiaochen	INFO	Train Epoch: 6 [26%]
2025-07-24 12:19:42,864	xiaochen	INFO	[2.898564338684082, 1.9539498090744019, 10.601505279541016, 14.936346054077148, 0.0, 1.3184711933135986, 100, 9.99250234335941e-05]
2025-07-24 12:19:51,176	xiaochen	INFO	====> Epoch: 6
2025-07-24 12:20:04,383	xiaochen	INFO	====> Epoch: 7
2025-07-24 12:20:17,453	xiaochen	INFO	====> Epoch: 8
2025-07-24 12:20:30,578	xiaochen	INFO	====> Epoch: 9
2025-07-24 12:20:43,664	xiaochen	INFO	====> Epoch: 10
2025-07-24 12:20:51,724	xiaochen	INFO	Train Epoch: 11 [53%]
2025-07-24 12:20:51,724	xiaochen	INFO	[2.9414026737213135, 1.9060834646224976, 12.477777481079102, 16.79501724243164, 0.0, 1.3030052185058594, 200, 9.986258590528146e-05]
2025-07-24 12:20:57,145	xiaochen	INFO	====> Epoch: 11
2025-07-24 12:21:10,073	xiaochen	INFO	====> Epoch: 12
2025-07-24 12:21:23,008	xiaochen	INFO	====> Epoch: 13
2025-07-24 12:21:35,927	xiaochen	INFO	====> Epoch: 14
2025-07-24 12:21:49,032	xiaochen	INFO	====> Epoch: 15
2025-07-24 12:22:00,284	xiaochen	INFO	Train Epoch: 16 [79%]
2025-07-24 12:22:00,284	xiaochen	INFO	[2.6509132385253906, 2.086803674697876, 12.517009735107422, 17.605146408081055, 0.0, 1.5448822975158691, 300, 9.980018739066937e-05]
2025-07-24 12:22:02,492	xiaochen	INFO	====> Epoch: 16
2025-07-24 12:22:15,650	xiaochen	INFO	====> Epoch: 17
2025-07-24 12:22:29,068	xiaochen	INFO	====> Epoch: 18
2025-07-24 12:22:42,137	xiaochen	INFO	====> Epoch: 19
2025-07-24 12:22:55,045	xiaochen	INFO	Saving model and optimizer state at iteration 20 to logs/xiaochen/logs_s2\G_233333333333.pth
2025-07-24 12:22:55,867	xiaochen	INFO	Saving model and optimizer state at iteration 20 to logs/xiaochen/logs_s2\D_233333333333.pth
2025-07-24 12:22:57,572	xiaochen	INFO	saving ckpt xiaochen_e20:Success.
2025-07-24 12:22:57,573	xiaochen	INFO	====> Epoch: 20