[Subject Prev][Subject Next][Thread Prev][Thread Next][Date Index][Thread Index]

[hts-users:02609] Re: About training in the STRAIGHT version HTS 2.1.1 script


Hi,
I have a question: Have you run straight version hts2.1.1beta completely?
Thanks
-Ali

2010/6/28 康世胤 <kangshiyin@xxxxxxxxx>
Hi,

No, I never complete the straight version hts2.1.1 demo with any configurations.
I give up and downgrade to hts2.1.1beta and hts2.1 now.



在 2010年6月28日 上午11:54,evan huang <vinsanity1101@xxxxxxxxx>写道:

Hello,

Thanks for your suggestion. I've used the StraightV40pcode and the STRAIGHT-version HTS 2.1.1 English demo and no modification is made.
The demo runs successfully until the step "Start generating speech parameter sequences (stc) ".
In this step, It seems that something wrong with the parameter generation, the warning shows as:
"WARNING [-9999]  OutProb: output prob at frame 4 is 0.0 in /usr/local/HTS-2.1.1/bin/HMGenS"
and then in the end, the log shows the ERROR and script stops.
All I can get is the 1mix synthesizes speech voices, and this result is as same as my previous results using STRAIGHTV40_007c.
So, Mr. Kang, do you successfully complete the original script using the STRAGHT_pcode?

Best Regards,
Yi-chin


2010/6/27 康世胤 <kangshiyin@xxxxxxxxx>

Hi,

Have you tried to use the default version "straightV40pcode" specified in the file INSTALL, without modifying the scripts?



2010/6/23 evan huang <vinsanity1101@xxxxxxxxx>

Hello,

I've downloaded the STRAIGHT version English demo script from the website and tried to train a Chinese acoustic model.
the script have been modified in order to use straightV40_007c (from 1024 to 2048, which is suggested from the mailing list) and Chinese dataset.
However, the training process stops at the "generating speech parameter sequences (stc)" and the error is showed as below:
------------------------------------------------------------------
...
 WARNING [-9999]  OutProb: output prob at frame 80 is 0.0 in /usr/local/HTS-2.1.1/bin/HMGenS
 WARNING [-9999]  OutProb: output prob at frame 81 is 0.0 in /usr/local/HTS-2.1.1/bin/HMGenS
 WARNING [-9999]  OutProb: output prob at frame 82 is 0.0 in /usr/local/HTS-2.1.1/bin/HMGenS
 WARNING [-9999]  OutProb: output prob at frame 83 is 0.0 in /usr/local/HTS-2.1.1/bin/HMGenS
.....
WARNING [-9999]  ParamGen: failed to compute output prob in /usr/local/HTS-2.1.1/bin/HMGenS

.....

Pruning-On[1500.0 100.0 5000.0]
 Generating Label sen_0001.lab
 Total number of frames = 1377
  PdfStream[1]: 1377 frames
  PdfStream[2]: 793 frames
  PdfStream[3]: 1377 frames
 EM-based parameter generation (Hidden: MIX, STATE)
 WARNING [-7324]  StepBack: Bad data or over pruning
 in /usr/local/HTS-2.1.1/bin/HMGenS
 WARNING [-9999]  ParamGen: failed to compute output prob in //usr/local/HTS-2.1.1/bin/HMGenS
  Iteration 1: Average LogP = -7.262164e+06
...
-----------------------------------------------------------

Although the 1mix model can be trained, and the model can be used to generate synthesized speech.
But why the stc step has such an error?
I've used the non-straight version and it can complete the all training procedures without error using the same dataset.
Maybe it is due the STRAIGHT version?
Could anyone suggest that the suitable version of STRAIGHT that I should use? Or other useful suggestions such as the parameter setting?
Thanks very much in advance for any kinds of help!

Best Regards,

YC Huang




--
康世胤
Shiyin Kang





--
康世胤
Shiyin Kang



References
[hts-users:02530] About training in the STRAIGHT version HTS 2.1.1 script, evan huang
[hts-users:02531] Re: About training in the STRAIGHT version HTS 2.1.1 script, 康世胤
[hts-users:02532] Re: About training in the STRAIGHT version HTS 2.1.1 script, evan huang
[hts-users:02533] Re: About training in the STRAIGHT version HTS 2.1.1 script, 康世胤