The numbers of the audio samples and speakers mismatch of LibriTTS dataset

ming024 / FastSpeech2

An implementation of Microsoft's "FastSpeech 2: Fast and High-Quality End-to-End Text to Speech"

MIT License

1.69k stars 515 forks source link

Open shiyanpei0826 opened 7 months ago

shiyanpei0826 commented 7 months ago

Your project and implementation is fantastic, I am learning our implementations

However, when I download your alignment Textgrids of Libritts, I found the numbers are much less than that of original Libritts dataset

Can you please share the details of alignments?

Thank you.