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

CV3-Eval

Benchmarks

Task NameDataset NameSOTA ResultTrend
Voice CloningCV3-Eval multilingual voice cloning (test)
WER (zh)2.65
18
Text-to-SpeechCV3-Eval Monolingual (hard-en)
WER4.37
7
Text-to-SpeechCV3-Eval Monolingual (hard-zh)
WER (%)8.55
7
Text-to-SpeechCV3-Eval Monolingual zh
WER2.65
7
Speech GenerationCV3-Eval ko
CER3.9
7
Multilingual Voice CloningCV3-Eval Multilingual Voice Cloning (hard-en)
WER5.93
6
Text-to-SpeechCV3-Eval Cross-lingual zh → en
WER4.32
5
Text-to-SpeechCV3-Eval Cross-lingual en → zh
WER8.01
5
Zero-shot Text-to-SpeechCV3-Eval Multilingual Voice Cloning
English CER/WER4.87
5
Zero-shot Voice CloningCV3-Eval
CV3-Eval Score (zh)4.08
4
Zero-Shot TTSCV3-Eval Cross-Lingual Zero-Shot (to-ko) (test)
CER (English Prompt)6
4
Zero-Shot TTSCV3-Eval Cross-Lingual Zero-Shot (to-ja) (test)
CER (English Prompt)14.12
4
Zero-Shot TTSCV3-Eval Cross-Lingual Zero-Shot (to-en) (test)
WER (zh)5.07
4
Zero-Shot TTSCV3-Eval Cross-Lingual Zero-Shot (to-zh) (test)
CER (English Prompt)7.16
4
Showing 14 of 14 rows