Instruction-following Text-to-Speech on OV-Speech
75.43Gemini ScoreGroundTruth
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| GroundTruth2026.01 | 75.43 | — | 3.1 | — | 4.1 | 4.33 | |
| OV-InstructTTS-TEP2026.01 | 70.42 | — | 3.61 | 0.722 | 4.28 | 3.91 | |
| GPT4oTimbre conversion=true2026.01 | 68.31 | — | 3.89 | 0.701 | 3.23 | 2.42 | |
| Step-Audio-2-mini2026.01 | 67.59 | — | 5.49 | 0.701 | 3.53 | 2.4 | |
| Cosyvoice2Instruction=No-Instruct2026.01 | 66.99 | — | 3.09 | 0.659 | 3.84 | 2.94 | |
| Higgs Audio V2Timbre conversion=true2026.01 | 65.1 | — | 8.42 | 0.707 | 3.81 | 3 |