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

[hts-users:01736] Re:


> I am trying to train new voice using HTS-demo. I recorded my voice used
> Praat software, mono channel, 16KHz,saved the recordings as wav files, and
> I used sox -c 1 -s -w -t wav -r 16000 $wav -c 1 -s -w -t raw -r 16000 $raw
> converted wav files to raw files. Because I used CMU-ARCTIC prompts, so I
> don't need build utts. when I run HTS-demo, make data, lfo, mgc, gv, cmp
> files could not generated. Did any one experience the same problem, any
> solution and suggestion will be appreciated.
>

Hi,

So I assume that the only thing you did was to replace the original SLT
raw files with your raw files?

What is the error message?
Can you build the original SLT demo properly?

The original data/labels/full/* files contain duration information which
corresponds with the original raw files, so they are now partially screwed
up. However, I don't think that's the problem here as the segmentation is
redone anyway. (Too busy to check this now, though.)

br,
  Nicholas



References
[hts-users:01735], Zheng, Dang C.