ConceptZero-shot Voice Conversion / 零样本VC
Zero-Shot Voice Conversion
Zero-shot voice conversion refers to voice conversion in which a model can convert speech into the timbre of a specified speaker without having seen data from that target speaker. Its core lies in disentangling content from speaker identity, and its generalization capability depends heavily on the language coverage of the training data.
Timeline (last 90 days)
Nov 3
Existing zero-shot voice conversion systems have clear deficiencies in pronunciation preservation for Arabic speech: converted speech maintains speaker similarity but degrades significantly in intelligibility and pronunciation accuracy.
Unverified50%