Audio Aesthetics Evaluation on Audiobox Aesthetics (AES) Speech
0.564PCC (CE)AES
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| AESevaluation_mode=Original code and weights2026.05 | 0.564 | 0.614 | 0.59 | 0.73 | 0.536 | 0.574 | 0.401 | 0.665 | |
| JASTINevaluation_mode=Original code and weights2026.05 | 0.531 | 0.594 | 0.601 | 0.707 | 0.497 | 0.561 | 0.441 | 0.666 | |
| NISQAevaluation_mode=Original code and weights2026.05 | 0.311 | 0.366 | 0.223 | 0.433 | 0.29 | 0.351 | 0.186 | 0.412 | |
| Gemini-3-Proevaluation_mode=API2026.05 | 0.307 | 0.216 | 0.429 | 0.381 | 0.279 | 0.206 | 0.292 | 0.386 | |
| UTMOSevaluation_mode=Original code and weights2026.05 | 0.286 | 0.306 | -0.148 | 0.36 | 0.26 | 0.273 | -0.148 | 0.306 | |
| Gemini-2.5-Proevaluation_mode=API2026.05 | 0.251 | 0.31 | 0.371 | 0.413 | 0.243 | 0.265 | 0.285 | 0.421 | |
| Gemini-2.5-Flashevaluation_mode=API2026.05 | 0.215 | 0.203 | 0.304 | 0.328 | 0.192 | 0.217 | 0.256 | 0.304 | |
| Qwen3-Omnievaluation_mode=Original code and weights2026.05 | 0.19 | 0.248 | 0.183 | 0.349 | 0.168 | 0.219 | 0.174 | 0.307 | |
| AudioFlamingo3evaluation_mode=Original code and weights2026.05 | 0.037 | 0.018 | 0.018 | 0.087 | 0.029 | 0.038 | 0.056 | 0.096 | |
| Qwen2-Audioevaluation_mode=Original code and weights2026.05 | 0.032 | -0.02 | -0.014 | -0.003 | 0.057 | 0.036 | -0.006 | 0.015 |