babysor / MockingBird

🚀AI拟声: 5秒内克隆您的声音并生成任意语音内容 Clone a voice in 5 seconds to generate arbitrary speech in real-time
Other
34.89k stars 5.18k forks source link

分享两个训练好的synthesizer模型 #289

Open wrk226 opened 2 years ago

wrk226 commented 2 years ago

俩模型都是用最新的代码来训练的,不需要切换回0.0.1 第一个模型synthesizer-merged_110k,是在代码支持的四个数据集(aidatatang_200zh,magicdata,aishell3,data_aishell)上联合训练的。learning rate=0.001无衰减,batch size=128,iteration=110k。 第二个模型synthesizer-zhvoice_170k,是在zhvoice这个数据集上训练的。learning rate=0.001无衰减,batch size=128,iteration=170k。 两个模型我都已经测试过是可用的,不过第一个似乎比第二个好一点。我觉得问题是出在vocoder上,因为我现在用的vocoder(wavernn)并不是在zhvoice数据集上训练的,而我也懒得再训练一个vocoder。不过用hifigan的话效果倒是差不多,不过音色又有明显不同,挺有意思的。 关于训练,我这俩模型还是可以进一步优化的,现在第一个模型的loss是在0.24左右,第二个是在0.22左右,不过花的时间太久我就懒得继续训练了。 下面是下载链接: 百度云 密码: ir90 Google Drive

xinranxi commented 2 years ago

赞啊,感谢感谢,家里破电脑没法训练,感谢模型

amzsowhat commented 2 years ago

感谢大佬!

babysor commented 2 years ago

手动 👍

Battle-l commented 2 years ago

感谢感谢

joshua54321 commented 2 years ago

大佬您好,在从openslr下载了magicdata数据集后,是否遇到过train_set.tar.gz解压失败的情况呢,最后报:gzip: stdin: invalid compressed data--format violated,确认是下载完了的。。

wrk226 commented 2 years ago

gzip: stdin: invalid compressed data--format violated

解压时候没遇到任何问题

flysmart commented 2 years ago

感谢分享!

SunchinSekian commented 2 years ago

感谢大佬,这个模型非常好

skyflym commented 2 years ago

可以用在现在最新的main分支上接着跑训练吗? 谢谢

x18-1 commented 2 years ago

新模型,新气象。感谢大佬

EbenZhang commented 1 year ago

我用main branch+这个synthesizer, 结果全是杂音/电流音 用别的 synthesizer 是好的. 请问大佬是不是我的打开方式不对?

以下是console输出:

Loaded encoder "pretrained.pt" trained to step 1594501
Synthesizer using device: cpu
Building hifigan
Loading 'vocoder\saved_models\pretrained\g_hifigan.pt'
Complete.
Removing weight norm...
Trainable Parameters: 0.000M
Loaded synthesizer "synthesizer-zhvoice_170k.pt" trained to step 170000
+----------+---+
| Tacotron | r |
+----------+---+
|   170k   | 2 |
+----------+---+

Read ['欢迎使用工具箱', '现已支持中文输入']
Synthesizing ['huan1 ying2 shi3 yong4 gong1 ju4 xiang1', 'xian4 yi3 zhi1 chi2 zhong1 wen2 shu1 ru4']

| Generating 1/1

Done.
silverLXT commented 9 months ago

hi,我也用main branch 和 这个synthesizer,结果也全是杂音,各位大佬这是为啥呀

Loaded synthesizer models: 2 Loaded encoders models: 1 Loaded vocoders models: 2 Loaded encoder "pretrained.pt" trained to step 1594501 Synthesizer using device: cpu Building hifigan Loading 'data/ckpt/vocoder/pretrained/g_hifigan.pt' Complete. Removing weight norm... Trainable Parameters: 0.000M Loaded synthesizer "synthesizer-merged_110k.pt" trained to step 110000 +----------+---+ | Tacotron | r | +----------+---+ | 110k | 2 | +----------+---+

Read ['欢迎使用工具箱', '现已支持中文输入'] Synthesizing ['huan1 ying2 shi3 yong4 gong1 ju4 xiang1', 'xian4 yi3 zhi1 chi2 zhong1 wen2 shu1 ru4']

| Generating 1/1

Done.

mikasa999 commented 7 months ago

还未下载,先谢为敬