Current inferencing scripts rename untranscribed audio files to audio.wav. The linked audio file in the Elan output file is a rel path to where the elan file was created in the Kaldi dir. Opening the generated Elan output file requires manually relinking audio.
Would be better if the audio file name was kept throughout, to save relinking files when reviewing the output.
The changes in this PR percolate the audio filename through the process, using it in the Elan linked media field, and naming the text and elan files with the same basename.
Leaving the ESPnet engine as using audio.wav. Hopefully the HFT work will be done soon and ESPnet removed. Not worth the bother of trying to get it running to check any changes to that engine.
Current inferencing scripts rename untranscribed audio files to
audio.wav
. The linked audio file in the Elan output file is a rel path to where the elan file was created in the Kaldi dir. Opening the generated Elan output file requires manually relinking audio.Would be better if the audio file name was kept throughout, to save relinking files when reviewing the output.
The changes in this PR percolate the audio filename through the process, using it in the Elan linked media field, and naming the text and elan files with the same basename.
Leaving the ESPnet engine as using
audio.wav
. Hopefully the HFT work will be done soon and ESPnet removed. Not worth the bother of trying to get it running to check any changes to that engine.