Our new paper, "Textless Dependency Parsing by Labeled Sequence Prediction" (
#interspeech2024
) is out!! Text-based cascades lose the prosody of speech, but our method avoids this! Nice work by Kando-san of Miyao lab, UTokyo.
New paper at
#interspeech2024
! Does your voice agent look at its surroundings and speak? We propose a new task "egocentric speech synthesis" and a method for building the corpus.
Paper:
Data:
we are thrilled to announce YODAS v2!
- 400k hours, 149 languages of speech data (same to v1)
- supporting long-form speech
- higher sampling rate (24 kHz sampling)
we are thrilled to announce YODAS v2!
- 400k hours, 149 languages of speech data (same to v1)
- supporting long-form speech
- higher sampling rate (24 kHz sampling)