Librispeech and Buckeye train infer with bad case

[panxin801]

Hello, and thank you for your sharing. This time I use Librispeech and Buckeye for training and 20 dataset from ljspeech for inference but i find that the time align is not monotony.

intervals [17]:
				xmin = 7.24
				xmax = 7.35
				text = "on"
			intervals [18]:
				xmin = 7.36
				xmax = 7.48
				text = "the"
			intervals [19]:
				xmin = 7.48
				xmax = 7.26
				text = "assassination"
			intervals [20]:
				xmin = 7.71
				xmax = 7.93
				text = "of"

[panxin801]

you can see the time align is not right, do you face the same problem like me in the experiment, and do you have some advice for this . Thank you again.