Voice Cloning And Speech Synthesis With Mouth Movement Regeneration

via “voice cloning and speaker adaptation”

text-to-speech model by undefined. 12,14,937 downloads.

Unique: Combines speaker-agnostic phonetic encoding with adaptive layer normalization in the decoder, enabling voice cloning from minimal reference audio without speaker-specific fine-tuning, while maintaining language-agnostic synthesis capabilities

vs others: Achieves voice cloning with shorter reference samples (3-5 seconds vs. 10-30 seconds for Glow-TTS variants) and maintains multilingual support simultaneously, unlike single-language voice cloning models