Project Icon

wespeaker

Comprehensive Solutions for Speaker Embedding and Verification

Product DescriptionThe wespeaker project offers a toolkit specializing in speaker embedding learning, aimed at facilitating speaker verification processes. It features online and kaldi-format feature extraction, and supports tasks including embedding, similarity computation, and diarization. Updated frameworks like Whisper-PMFA and ssl pre-trained models improve usability. This toolkit is beneficial for research and production by enhancing performance through methods like score calibration.
Project Details