Does NeMo data manifest format support long audio?

[feddybear]

As the title suggests, does the data manifest only assume that each line starts from the beginning of the file referred to by audio_filepath?
Is it not possible to have a long recording where you know the utterance boundaries and use the same file, indicate both start time and duration?

Sorry if I haven't spent much time finding this information from all documentations, just starting to ask from going over the tutorial

[titu1994]

Yes it supports it. You can use "offset" and "duration" keys in seconds in order to select a subset of the audio file.

[titu1994]

However you will need multiple rows for the same audio file to select different subsegments

[feddybear]

Yes, no problem with that. I'm just glad I don't have to trim them. Thanks!!!