Weird Audio Generated

Archives TTS (Text-to-Speech)
#1

Hi there! Would like to ask a question regarding the issue on the title.

Before that, below is my dataset specification:

° a total of 5 hours dataset
° contain >30 seconds for most of the audios

When the model is ready, i tried to test it with few words that has the highest frequency in my corpus:

Screenshot_20190709-172937_Excel
Screenshot_20190709-172937_Excel.jpg612×1333 212 KB

I picked the highest word ‘saya’ and the audio generated very long for only 1 word.

Only the first 1 seconds can hear the word ‘saya’ and the next few seconds became unknown sound.

Also during testing, i was getting below error:

IMG-20190709-WA0013
IMG-20190709-WA0013.jpg1080×607 97 KB

(Stopped with decoder…)

2 question. What could be the reason of me having that kind of mentioned audio ?

AND

is the stopped decoder issue due to long sentence audio (dataset) ?

#2

your attention or stopnet does not work properly so network reaches the maximum iteration.

#3

I’m sorry but what caused for the issue you mentioned ya ?

#4

cannot tell without knowing more. You need to find out yourself. Ya!