forced-alignment

Tools for forced alignment. Fork of p2fa.

What's here ----------- 1. [`create_transcripts.py`](https://github.com/myedibleenso/forced-alignment/blob/master/transcripts/create_transcripts.py)

An original script for automatic generation of **htk**-compatible transcripts from some set of source sentences

2. pronunciator.py

   Pronunciation estimator for unknown utterances (requires a pronunciation dictionary).  A copy of the CMU pronouncing dictionary is provided here. This script is able to generate .csv and LaTeX-formatted table of compositional statistics for a specified transcript or set of sentences (i.e. average length, # syllables, syllables per word, unique phones, etc).

3. align.py

   Minimally modified p2fa that resolves deprecated SoX call

4. align_batch.py

   Script to batch process multiple wav files with minimally modified p2fa.

5. align_new.py

   Revised p2fa with several new features

6. splice.praat

   Praat script for TextGrid-based slicing of wav files.  Credit and thanks given to Jessamyn Schertz

Setup

Install htk 3.4. Careful about this point as v3.4.1 will not work.

Mac users experiencing problems should look here.

Other stuff -----------

File naming convention:

<GENDER><GENDER-ID>_HS_<NATIVE or NON-NATIVE (N or NN)>_<SPEED (NORM or FAST)>_<LIST>-<SENTENCE NUM (1-10)>

MORE RECORDINGS: http://www.voiptroubleshooter.com/open_speech/american.html