Support for VALL-E X

[152334H]

https://vallex-demo.github.io/

Title

[takan1]

@lifeiteng Thanks for the great work! Any plan to implement vallex?

[lifeiteng]

hands on it now.

[Li-JEN]

Hi, @lifeiteng I have tried the vall-ex based on your code. But the modification only on adding the language ID and using the BigCiDian mentioned in paper as phoneme input.

I utilized LibriTTS and AISHELL3 as data, and below is the preliminary result. Chinese speaker Audio prompt: link

Text: 托管费在基金管理费之外征收 Synthesized Audio: link Text: This I read with great attention while they sat silent. Synthesized Audio: link

I am not sure if I do it correctly, Maybe I can discuss it with you? This is my email: mail

[misakaikato]

Hi, @lifeiteng I have tried the vall-ex based on your code. But the modification only on adding the language ID and using the BigCiDian mentioned in paper as phoneme input.

I utilized LibriTTS and AISHELL3 as data, and below is the preliminary result. Chinese speaker Audio prompt: link

Text: 托管费在基金管理费之外征收 Synthesized Audio: link Text: This I read with great attention while they sat silent. Synthesized Audio: link

I am not sure if I do it correctly, Maybe I can discuss it with you? This is my email: mail

hi, the link to the first audio you synthesized using the prompt seems to be wrong, and now it's still that prompt.

朋友，中文合成的第一条链接好像是错误的，打开还是那个prompt的文件。😂

[thomasbohm]

@lifeiteng Hey thanks for the efforts! Any update on VALL-E X?

[skysbird]

Hi, @lifeiteng I have tried the vall-ex based on your code. But the modification only on adding the language ID and using the BigCiDian mentioned in paper as phoneme input.

I utilized LibriTTS and AISHELL3 as data, and below is the preliminary result. Chinese speaker Audio prompt: link

Text: 托管费在基金管理费之外征收 Synthesized Audio: link Text: This I read with great attention while they sat silent. Synthesized Audio: link

I am not sure if I do it correctly, Maybe I can discuss it with you? This is my email: mail

same to me, my solution is same as you. it seems works

[RahulBhalley]

Hi, @lifeiteng I have tried the vall-ex based on your code. But the modification only on adding the language ID and using the BigCiDian mentioned in paper as phoneme input. I utilized LibriTTS and AISHELL3 as data, and below is the preliminary result. Chinese speaker Audio prompt: link Text: 托管费在基金管理费之外征收 Synthesized Audio: link Text: This I read with great attention while they sat silent. Synthesized Audio: link I am not sure if I do it correctly, Maybe I can discuss it with you? This is my email: mail

same to me, my solution is same as you. it seems works

अपने VALLE-X को कैसे इंप्लीमेंट किया?

[skysbird]

maybe you can refer to my e-x branch forked from here

[RahulBhalley]

आपके रेपॉजिटरी में VALLE-X से संबंधित कोई कोड अपडेट दिखाई नहीं दे रहा है। @skysbird

[Li-JEN]

maybe you can refer to my e-x branch forked from here

@skysbird How do you handle the part of Language ID? I saw your code, but it seems like you follow the architecture of origin Vall-E?

[skysbird]

maybe you can refer to my e-x branch forked from here

@skysbird How do you handle the part of Language ID? I saw your code, but it seems like you follow the architecture of origin Vall-E?

I have embedded language id

[RahulBhalley]

maybe you can refer to my e-x branch forked from here

@skysbird How do you handle the part of Language ID? I saw your code, but it seems like you follow the architecture of origin Vall-E?

I have embedded language id

Can you please drop a link to that code statement?

Nvm, found it.

[treya-lin]

So is there any update on implementing VALL-E X?