gpt-sovits模型《恋与深空》秦彻声音配音模型

plxg

19小时前更新

370

该帖子部分内容已隐藏

角色介绍

秦彻是《恋与深空》中的第四位男主角，他的角色设定充满了神秘和危险的气息。他是神秘组织“暗河”的首领，这个组织在游戏中几乎无所不做，而秦彻作为首领，自然也是“臭名昭著”的。他的外表特征包括白毛红瞳，展现出一种浓颜的帅气。

在游戏剧情中，秦彻对玩家扮演的女主角进行了持续的关注和监视，这种设定让他的角色更加神秘和吸引人。此外，他的配音演员包括张磊、桑毓泽和小西克幸，这些配音演员的出色表现进一步增强了角色的魅力。

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。

模型训练日志

Variable._execution_engine.run_backward(  # Calls into the C++ engine to run the backward pass
INFO:peiyin.me_qinche:Train Epoch: 1 [0%]
INFO:peiyin.me_qinche:[2.3872570991516113, 2.1239542961120605, 8.522953987121582, 20.069467544555664, 0.0, 1.7259124517440796, 0, 9.99875e-05]
12it [00:26,  2.18s/it]
INFO:peiyin.me_qinche:====> Epoch: 1
12it [00:19,  1.66s/it]
INFO:peiyin.me_qinche:====> Epoch: 2
12it [00:19,  1.62s/it]
INFO:peiyin.me_qinche:====> Epoch: 3
12it [00:14,  1.23s/it]
INFO:peiyin.me_qinche:====> Epoch: 4
12it [00:09,  1.33it/s]
INFO:peiyin.me_qinche:====> Epoch: 5
12it [00:09,  1.33it/s]
INFO:peiyin.me_qinche:====> Epoch: 6
12it [00:08,  1.34it/s]
INFO:peiyin.me_qinche:====> Epoch: 7
12it [00:09,  1.28it/s]
INFO:peiyin.me_qinche:====> Epoch: 8
4it [00:07,  1.77s/it]INFO:peiyin.me_qinche:Train Epoch: 9 [33%]
INFO:peiyin.me_qinche:[2.5681345462799072, 2.2206671237945557, 7.051852703094482, 18.716594696044922, 0.0, 1.4125581979751587, 100, 9.98875562335968e-05]
12it [00:14,  1.18s/it]
INFO:peiyin.me_qinche:====> Epoch: 9
12it [00:09,  1.21it/s]
INFO:peiyin.me_qinche:====> Epoch: 10
12it [00:15,  1.26s/it]
INFO:peiyin.me_qinche:====> Epoch: 11
12it [00:09,  1.31it/s]
INFO:peiyin.me_qinche:====> Epoch: 12
12it [00:09,  1.29it/s]
INFO:peiyin.me_qinche:====> Epoch: 13
12it [00:09,  1.33it/s]
INFO:peiyin.me_qinche:====> Epoch: 14
12it [00:09,  1.32it/s]
INFO:peiyin.me_qinche:Saving model and optimizer state at iteration 15 to logs/peiyin.me_qinche/logs_s2\G_233333333333.pth
INFO:peiyin.me_qinche:Saving model and optimizer state at iteration 15 to logs/peiyin.me_qinche/logs_s2\D_233333333333.pth
INFO:peiyin.me_qinche:saving ckpt peiyin.me_qinche_e15:Success.
INFO:peiyin.me_qinche:====> Epoch: 15
"H:\图\GPT-SoVITS-beta0217fix3\GPT-SoVITS-beta0217fix3\runtime\python.exe" GPT_SoVITS/s1_train.py --config_file "H:\图\GPT-SoVITS-beta0217fix3\GPT-SoVITS-beta0217fix3\TEMP/tmp_s1.yaml"
Seed set to 1234
Using 16bit Automatic Mixed Precision (AMP)
GPU available: True (cuda), used: True
TPU available: False, using: 0 TPU cores
IPU available: False, using: 0 IPUs
HPU available: False, using: 0 HPUs
<All keys matched successfully>
ckpt_path: None
[rank: 0] Seed set to 1234
Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/1
----------------------------------------------------------------------------------------------------
distributed_backend=gloo
All distributed processes registered. Starting with 1 processes
----------------------------------------------------------------------------------------------------

Missing logger folder: logs\peiyin.me_qinche\logs_s1\logs_s1
semantic_data_len: 83
phoneme_data_len: 83
        item_name                                     semantic_audio
0    qinche_1.wav  54 146 953 338 490 490 545 405 540 215 704 775...
1   qinche_11.wav  54 274 714 23 334 750 592 186 224 638 76 391 1...
2   qinche_13.wav  1012 274 98 490 721 214 490 490 490 32 490 490...
3   qinche_15.wav  721 29 162 98 318 382 807 836 914 563 199 481 ...
4   qinche_17.wav  54 234 875 505 505 875 41 65 809 875 234 170 6...
..            ...                                                ...
78  qinche_76.wav  54 234 540 809 875 234 721 290 190 775 823 914...
79  qinche_78.wav  520 262 260 170 721 721 98 98 98 98 721 98 17 ...
80   qinche_8.wav  1005 714 360 14 420 551 962 502 690 1016 380 8...
81  qinche_81.wav  54 170 505 234 41 29 29 809 875 875 875 721 72...
82  qinche_83.wav  721 524 274 721 490 490 490 490 545 668 145 37...

[83 rows x 2 columns]
deleted 5 audios who's phoneme/sec are bigger than 25 or smaller than 3
dataset.__len__(): 156
LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0]

  | Name  | Type                 | Params
-----------------------------------------------
0 | model | Text2SemanticDecoder | 77.5 M
-----------------------------------------------
77.5 M    Trainable params
0         Non-trainable params
77.5 M    Total params
309.975   Total estimated model params size (MB)
H:\图\GPT-SoVITS-beta0217fix3\GPT-SoVITS-beta0217fix3\runtime\lib\site-packages\pytorch_lightning\loops\fit_loop.py:293: The number of training batches (15) is smaller than the logging interval Trainer(log_every_n_steps=50). Set a lower value for log_every_n_steps if you want to see logs for the training epoch.
Epoch 14: 100%|█| 15/15 [00:01<00:00,  8.55it/s, v_num=0, total_loss_step=428.0, lr_step=0.002, top_3_acc_step=0.677, t`Trainer.fit` stopped: `max_epochs=15` reached.
Epoch 14: 100%|█| 15/15 [00:09<00:00,  1.56it/s, v_num=0, total_loss_step=428.0, lr_step=0.002, top_3_acc_step=0.677, t
"H:\图\GPT-SoVITS-beta0217fix3\GPT-SoVITS-beta0217fix3\runtime\python.exe" GPT_SoVITS/inference_webui.py
<All keys matched successfully>
Number of parameter: 77.49M
DEBUG:root:Using proactor: IocpProactor
DEBUG:root:Using proactor: IocpProactor
Running on local URL:  http://0.0.0.0:19872
实际输入的参考文本: 我出门不看天气预报，会被天气影响的计划一般都不重要。
实际输入的目标文本: 看清楚了么？这才是你内心真正的欲望。秦彻是菲罗斯星有史以来悬赏金额最高的罪犯，也是唯 一一个从时空监狱里越狱成功的传说。越狱后以一己之力将穹羽星星主收服的妥妥帖帖，让星主取消星际关卡的警戒，并且开始对堡垒里的飞船进行清点，彻底掌控了穹羽星的力量，并在各领主提供的请报下得知“以太芯核”、“以太之眼”的下落，得到想要的情报后在离开穹羽星之际使用Evol将整个堡垒尽数摧毁，穹羽星自星主到全体领主无一人生还。
实际输入的目标文本(切句后): 看清楚了么？这才是你内心真正的欲望
秦彻是菲罗斯星有史以来悬赏金额最高的罪犯，也是唯 一一个从时空监狱里越狱成功的传说
越狱后以一己之力将穹羽星星主收服的妥妥帖帖，让星主取消星际关卡的警戒，并且开始对堡垒里的飞船进行清点，彻底掌控了穹羽星的力量，并在各领主提供的请报下得知“以太芯核”、“以太之眼”的下落，得到想要的情报后在离开穹羽星之际使用Evol将整个堡垒 尽数摧毁，穹羽星自星主到全体领主无一人生还