Page Comparison

...

• Location: /tools/VisemeSchedulerSphinx4
• Language: C#Java
• Distribution: source
• Platform(s): Windows

...

This tool will do the following for each .wav file in the specified folder

where

- folder is a relative or absolute path to the folder containing the .wav and .txt files
- (default 'Ao') viseme to use is the single viseme it will use to map the audio to
- (default 6) scale factor is a multiply factor to the viseme weight data. If your results are too low or too high, you can adjust this value.
- (default 100) sample length is how much to split the audio file up for analyzing.

Run the VisemeSchedulerSox tool:

...

•  run Run the .wav through Sox to get the audio value at each sample rate
•  Map the value to a viseme weight
•  Process the .txt transcript of the audio and do an even distribution of the word break data based on the length of the audio file
• Write out the .bml file

...

• file through ForceAlignment (based on sphinx4 library) with the correct transcription (.txt file with the same name present in the same folder) and get a phoneme schedule with word break information (which is loaded in memory).
• Use the phoneme schedule, the word break data and the phoneme-to-viseme mapping to create a .bml file.
• Note that the viseme curve currently generated is trapezoidal (ramp-up and ramp-down for 10% of the beginning and end respectively).

Usage: process.bat <absolute path for folder which contains the .wav / .txt files>

where process.bat is located in /tools/VisemeSchedulerSphinx4

This will create a set of smartbody .bml files to use with your audio files in the folder specified.

FAQ

See the Main FAQ. Please use the Google Groups emailing list for unlisted questions.

...