How to Use MFA on the VCTK Dataset to Obtain .TextGrid Files。

[zhaojingxin123]

Which lexicon txt did you use to get TextGrid on vctk。
How to Use MFA on the VCTK Dataset to Obtain .TextGrid Files。

[li1jkdaw]

Hi, @zhaojingxin123 !
In our work we used MFA v1.0, but you can use its newer version MFA v2.0 (https://montreal-forced-aligner.readthedocs.io/en/latest/) which is more user friendly and has newer and better models.

To align any annotated English speech corpus, you can follow use case 3 from the guide ("You have a speech corpus, the language has a pretrained acoustic model and pretrained G2P model, but it doesn’t have a pretrained dictionary.") You can download pretrained acoustic and G2P models by the links, just make sure that you download compatible versions and the phoneset they use is ARPA (we used ARPABET in our work and didn't test our models on other phonesets). E.g. you can download English (US) ARPA acoustic and G2P models v2.0.0a.

[zhaojingxin123]

Hi, @zhaojingxin123 ! In our work we used MFA v1.0, but you can use its newer version MFA v2.0 (https://montreal-forced-aligner.readthedocs.io/en/latest/) which is more user friendly and has newer and better models.

To align any annotated English speech corpus, you can follow use case 3 from the guide ("You have a speech corpus, the language has a pretrained acoustic model and pretrained G2P model, but it doesn’t have a pretrained dictionary.") You can download pretrained acoustic and G2P models by the links, just make sure that you download compatible versions and the phoneset they use is ARPA (we used ARPABET in our work and didn't test our models on other phonesets). E.g. you can download English (US) ARPA acoustic and G2P models v2.0.0a.
thank you for your reply，happy Chinese newyear.