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

[hts-users:04509] Re: Speech Synthesis for New Language

Prepare for you:

1 . Recording data

2. Language model : 
- lexicon : word -> phonemes list , to create mono file
- question 
- tool, that convert txt to full-context label

Note : raw file is data of wav file
wave file == header + raw
you can view raw file with https://sourceforge.net/projects/wavesurfer/ (in demo raw file, choice Sample = 48000Hz, SampleEncoding Lin16, chanel Mono
and you can hear that)

Final : questions + labels/full + labels/mono + raw
utt  is optional

2017-03-18 13:49 GMT+07:00 Atlas Khan <atlaskhan90@xxxxxxxxx>:

I want to do speech synthesis using HTS for Urdu Language. I have ran Demo Script for English (HTS-demo_CMU-ARCTIC-SLT.tar.bz2) and it is run also ning fine. I also have explored all files in data directory of above demo. There are following different type of files in data directory
  • questions: list of all context and properties format for tree-based context clustering.
  • labels: phone labeled file for xxx.raw with/without their context and properties
  • raw: I think they are recordings
  • txt: Text of  corresponding recordings
  • utts: Utterance files
What I wanted to ask is which of these types of data I needed to prepare for Urdu language. I have speech corpus and their text labels. Kindly also tell me how to prepare this data. 

[hts-users:04508] Speech Synthesis for New Language, Atlas Khan