Attention weights from neural machine translation are used to transfer prosodic phrase boundaries from the source to the target language, and the resulting phrase durations condition a text-to-speech system to align dubbed speech with lip movements.
This is especially useful for online video streamers who publish often and lack access to professional translation services
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2019 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Prosodic Phrase Alignment for Machine Dubbing
Attention weights from neural machine translation are used to transfer prosodic phrase boundaries from the source to the target language, and the resulting phrase durations condition a text-to-speech system to align dubbed speech with lip movements.