Alignment
Forced Alignment is a process of aligning an audio recording to its transcription, which generates phone-level segmentation. Here you can submit an audio file (in wav or mp3 format) and its transcription as a text file. Please use the same filename (without its extension) for the audio file and the transcription, i.e: shakespeare.mp3 and shakespeare.txt. Please also provide an arbitrary project name which will be used later as a directory name.
You can also try following audio/text files for testing purpose:- English: shakespeare-01.mp3, shakespeare-01.txt
- Javanese: sekarkarya_01_dhangdhanggula_babad_jaka_tingkir.mp3, sekarkarya_01_dhangdhanggula_babad_jaka_tingkir.txt
Limitation
Currently, this web application can only force align audio files for up to 15 minutes due to network connection timeout. This limitation does not happen with our offline version to align longer audio files. We will improve the design of the web application using Message Queuing Protocol (such as RabbitMQ) to fix this limitation and to scale the back-end server horizontally.