Segment
#1
by
pltobing
- opened
Hi, thanks very much for the great work. This dataset project will bring a lot of varieties in the current "ideal TTS datasets" environment. We usually use the latter in all publications/research evaluations, comparisons, samples demo, etc., but they are not suitable for real-world production-level demands.
Regarding the data availability, are you also planning to segment the 24k audio as in the yodasv1 (16k)?
Many thanks for your confirmation.
hi, thanks for your encouraging words!
For your question, it is actually very easy to retrieve the segments for yodas2 (24k) since we have attached start, end timestamp for each segment (utterance) in this dataset