Controllable Text-to-Speech on Composite-instruction benchmark
0.841S-SIMAgentSteerTTS
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| AgentSteerTTSCategory=Ours, Backend=Gemini3 backend2026.05 | 0.841 | 1.34 | 0.955 | 3.82 | 4.28 | 4.4 | 4.38 | |
| CosyVoice2Category=Flow/Open2026.05 | 0.825 | 1.88 | 0.748 | 4.31 | 4.31 | 4.15 | 4.35 | |
| IndexTTS2Category=LLM-TTS2026.05 | 0.823 | 1.81 | 0.864 | 4.24 | 4.24 | 4.42 | 4.15 | |
| AgentSteerTTSCategory=Ours, Backend=Qwen3 backend2026.05 | 0.817 | 1.57 | 0.921 | 3.83 | 4.26 | 4.36 | 4.36 | |
| CosyVoiceCategory=Flow/Open2026.05 | 0.812 | 1.95 | 0.732 | 4.18 | 4.18 | 3.98 | 4.21 | |
| IndexTTSCategory=LLM-TTS2026.05 | 0.804 | 2.1 | 0.842 | 4.15 | 4.15 | 4.28 | 4.08 | |
| F5-TTSCategory=Flow/Open2026.05 | 0.795 | 2.14 | 0.715 | 4.05 | 4.05 | 3.92 | 4.15 | |
| Spark-TTSCategory=LLM-TTS2026.05 | 0.788 | 2.32 | 0.724 | 4.12 | 4.12 | 4.05 | 4.12 | |
| VALL-ECategory=Flow/Open2026.05 | 0.712 | 4.25 | 0.685 | 3.52 | 3.52 | 3.41 | 3.58 |