search

openvpi / DiffSinger

An advanced singing voice synthesis system with high fidelity, expressiveness, controllability and flexibility based on DiffSinger: Singing Voice Synthesis via Shallow Diffusion Mechanism
Apache License 2.0
2.69k stars 283 forks source link

Make pauses / silence inside singing #126

Closed MrDiplodocus closed 1 year ago

MrDiplodocus commented 1 year ago

Hi, Thank you for the great work! The results are really amazing! I trained the model and now I have some questions about pauses inside the sample. Is it possible to synthesize a single sample with the silence inside? Adding SP does not help, because the right syllable is concatenated and the model simply sings (lengthens) the syllable, and does not pause it. image

yqzhishen commented 1 year ago

What editor are you using, and if you use *.ds files to synthesize, how did you get/generate them? Adding SPs is expected to work.

MrDiplodocus commented 1 year ago

I don't use an editor. I create *.ds files programmatically. I looked in samples/ and it looks like I found an error with ph_num. I connected SP as AP according to 3. Add ph_num attribute from making of variance model. In the samples/ I saw that ph_num for SP is equal to 1. Thanks for the quick answer!