List your company

Text-to-Spatial Audio Generation on Spatial Audio Caption (test)

4.65MOS (Spatial Quality)Ground Truth
3.7143.9574.24.443May 29, 2026

Evaluation results

MethodLinks
Ground Truth2026.054.65——4.76
SwanSphereParams=1.09B, Inf. Time=0.21s/9.13s2026.054.31142.81.434.43
OmniAudio(text)Params=1.22B, Inf. Time=0.89s2026.054.11174.131.834.16
Tango2+ASParams=0.86B, Inf. Time=2.12s2026.053.95235.712.423.27
AudioLDM-2+ASParams=0.71B, Inf. Time=7.64s2026.053.86294.172.453.53
MMAudio+ASParams=1.03B, Inf. Time=2.76s2026.053.75313.262.773.44