Text-to-Audio Generation on AudioCaption (test)
0.526CLAP ScoreReference
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Reference2023.01 | 0.526 | — | — | 74.7 | 80.5 | |
| Make-An-AudioText-cond=T5-Large, Params=563M2023.01 | 0.486 | 4.83 | 2.81 | 71.8 | 77.2 | |
| Make-An-AudioText-cond=CLAP, Params=332M2023.01 | 0.482 | 4.61 | 2.79 | 72.5 | 78.6 | |
| Make-An-AudioText-cond=BERT, Params=809M2023.01 | 0.48 | 5.15 | 2.89 | 70.5 | 77.2 | |
| Make-An-AudioText-cond=CLIP, Params=576M2023.01 | 0.444 | 6.45 | 2.91 | 72.1 | 75.4 | |
| DiffsoundText-cond=CLIP, Params=520M2023.01 | 0.42 | 7.17 | 3.57 | 67.1 | 70.9 |