Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Emilia

Benchmarks

Task NameDataset NameSOTA ResultTrend
Speech CodingEmilia clean subset (test)
DNSMOS3.646
10
Text-to-SpeechEmilia EN speaking-rate
MUSHRA Score68.8
9
Speaker UnlearningEmilia (Seen opt-out set (-SO))
WER3.13
5
Semantic EvaluationEmilia (val)
PPL (Initial)8.42
4
Showing 4 of 4 rows