Text-to-Speech on 50 utterances (test)
97TTFP (ms)Qwen3-TTS
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Qwen3-TTSFrequency=12 Hz, Model Size=0.6B, Concurrency=1, Type=autoregressive2026.05 | 97 | 0.288 | |
| Qwen3-TTSFrequency=12 Hz, Model Size=1.7B, Concurrency=1, Type=autoregressive2026.05 | 101 | 0.313 | |
| Chatterbox-FlashFrequency=25 Hz, Model Size=0.5B, Block size (D)=32, Early decoding threshold (alpha)=0.75, Concurrency=12026.05 | 103 | 0.076 | |
| Chatterbox-FlashFrequency=25 Hz, Model Size=0.5B, Block size (D)=24, Early decoding threshold (alpha)=0.75, Concurrency=12026.05 | 105 | 0.084 | |
| Chatterbox-FlashFrequency=25 Hz, Model Size=0.5B, Block size (D)=16, Early decoding threshold (alpha)=0.75, Concurrency=12026.05 | 106 | 0.091 | |
| Chatterbox-FlashFrequency=25 Hz, Model Size=0.5B, Block size (D)=32, Early decoding threshold (alpha)=0.5, Concurrency=12026.05 | 115 | 0.09 | |
| Chatterbox-FlashFrequency=25 Hz, Model Size=0.5B, Block size (D)=16, Early decoding threshold (alpha)=0.5, Concurrency=12026.05 | 118 | 0.107 | |
| Chatterbox-FlashFrequency=25 Hz, Model Size=0.5B, Block size (D)=24, Early decoding threshold (alpha)=0.5, Concurrency=12026.05 | 119 | 0.1 | |
| Qwen3-TTSFrequency=25 Hz, Model Size=0.6B, Concurrency=1, Type=autoregressive2026.05 | 138 | 0.234 | |
| Qwen3-TTSFrequency=25 Hz, Model Size=1.7B, Concurrency=1, Type=autoregressive2026.05 | 150 | 0.253 |