search

begeekmyfriend / tacotron2

Forked from NVIDIA/tacotron2 and merged with Rayhane-mamah/Tacotron-2
BSD 3-Clause "New" or "Revised" License
81 stars 38 forks source link

segment a uttence #6

Closed freecui closed 4 years ago

freecui commented 4 years ago

is there any tool to segment a long uttence with aligning text on chinese?

begeekmyfriend commented 4 years ago

Add a punctuation such as comma among the pinyin transcript.

freecui commented 4 years ago

I don't have data with the pinyin transcript with a punctuation . I mean I want to use the long audios as traing data, but I just use short audios due to gpu memeory limit; so I want to segment audio with corresponding transcript, such as Montreal forced aligner

begeekmyfriend commented 4 years ago

As for audio clipping I suggest https://github.com/keithito/tacotron/issues/129. But the deficiency is that you need to match the transcript with the wav clip length rather than vise versa.

freecui commented 4 years ago

Is there any tool to add punctuation for a transcript

begeekmyfriend commented 4 years ago 
bash scripts/griffin_lim_synth.sh