BenchmarkSpeaker Embedding Cosine Similarity

SECS

An objective metric that computes the cosine similarity of embeddings derived from a speaker verification model, used to evaluate speaker similarity in voice cloning.

Timeline (last 90 days)

Oct 23

The industry commonly uses speaker similarity (such as SECS) and word error rate (WER) as objective evaluation metrics for zero-shot speech cloning.

Unverified50%

All Facts (1)

Unverified

The industry commonly uses speaker similarity (such as SECS) and word error rate (WER) as objective evaluation metrics for zero-shot speech cloning.

50%

Source Articles

VoxCPM2: Tokenizer-Free Speech Synthesis with New Breakthroughs in Multilingual Generation and CloningOct 23