The quality of automatic speech segmentation tools has been improved over the years and there are many proposals. Can you recommend a good free tool for this purpose?
I'm looking for something that can receive as input an audio file and provide a list of time instants as output.
I have the phonetic sequence and I don't need to have "perfect" segmentation boundaries.
There is DARLA's completely automated Vowel extraction: http://darla.dartmouth.edu/uploadasr it works best with American English though.
If you have a transcribed Praat textgrid you can use the semi automated version: http://darla.dartmouth.edu/uploadbound and that is very good.
Thank you Bradley Mackay ! It seems a nice tool and I'll give it a try.
Have you tried FAVE/FAV or MAUS?
FAVE is available both for download https://github.com/JoFrhwld/FAVE or you can use it via a web interface (fave.ling.upenn.edu which seems to be loading really slowly today). FAVE is based on American English, but if you have the phonetic transcription you can use it for things like Finnish (like I've done at some point).
MAUS provides more fitted languages and people seem to like it too.
https://www.en.phonetik.uni-muenchen.de/research/current_projects/maus_en.html
Montreal Forced Aligner is also a good option for speech segmentation.
https://montreal-forced-aligner.readthedocs.io/en/latest/
- Badges
- Science topic
- Similar topics
- Mathematics