Target-Speaker Multilingual Speech Generation on TTS multilingual Chinese (test)
70.8WERQwen3-TTS
Evaluation results
| Method | Links | |
|---|---|---|
| Qwen3-TTSTokenizer Frequency=25Hz, Parameters=1.7B, Evaluation Protocol=CustomVoice2026.01 | 70.8 | |
| Qwen3-TTSTokenizer Frequency=25Hz, Parameters=0.6B, Evaluation Protocol=CustomVoice2026.01 | 87.4 | |
| Qwen3-TTSTokenizer Frequency=12Hz, Parameters=1.7B, Evaluation Protocol=CustomVoice2026.01 | 90.3 | |
| Qwen3-TTSTokenizer Frequency=12Hz, Parameters=0.6B, Evaluation Protocol=CustomVoice2026.01 | 94.4 | |
| GPT-4o-Audio Preview2026.01 | 351.9 |