Style-controlled Speech Synthesis on Seed-TTS-eval en (test)
5.59SPS (Steps Per Second)Fast LoRA
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Fast LoRAStyle Control Axis=Speed-axis control, Operation=speed-up, Zero-shot transfer=true2026.06 | 5.59 | 120.4 | 191 | 3.49 | 0.617 | 3.3 | 4.72 | 4.68 | |
| DSP speed-upStyle Control Axis=Speed-axis control, Operation=speed-up, Zero-shot transfer=true2026.06 | 5.48 | 121.2 | 195.7 | 3.5 | 0.475 | 1.56 | 3.08 | 2.76 | |
| Low-pitch LoRAStyle Control Axis=Pitch-axis control, Operation=pitch-down, Zero-shot transfer=true2026.06 | 3.68 | 108.9 | 164.6 | 3.11 | 0.632 | 3.16 | 4.6 | 4.84 | |
| Baseline (CosyVoice2-0.5B)Style Control Axis=Baseline, Zero-shot transfer=true2026.06 | 3.65 | 120.4 | 192.2 | 2.81 | 0.655 | 3.28 | — | — | |
| DSP pitch-upStyle Control Axis=Pitch-axis control, Operation=pitch-up, Zero-shot transfer=true2026.06 | 3.65 | 150.9 | 239.9 | 2.59 | 0.173 | 1.57 | 1.4 | 2.28 | |
| DSP pitch-downStyle Control Axis=Pitch-axis control, Operation=pitch-down, Zero-shot transfer=true2026.06 | 3.65 | 98 | 155.4 | 3.28 | 0.158 | 1.49 | 1.4 | 2 | |
| High-pitch LoRAStyle Control Axis=Pitch-axis control, Operation=pitch-up, Zero-shot transfer=true2026.06 | 3.61 | 156.1 | 241 | 3.01 | 0.609 | 3.37 | 4.12 | 4.4 | |
| Slow LoRAStyle Control Axis=Speed-axis control, Operation=slow-down, Zero-shot transfer=true2026.06 | 2.3 | 122.1 | 194.3 | 3.18 | 0.65 | 3.05 | 4.56 | 4.24 | |
| DSP slow-downStyle Control Axis=Speed-axis control, Operation=slow-down, Zero-shot transfer=true2026.06 | 2.19 | 121.7 | 193.7 | 2.67 | 0.5 | 1.45 | 2.76 | 2.28 |