Text-to-audio generation on VGGSound benchmark leaderboard · SOTA2 Research

Text-to-audio generation on VGGSound

1.44Fréchet Audio Distance (FAD)AudioX-Turbo

Evaluation Results

Method	Links
AudioX-Turbo2026.06		1.44	—	—	—	—	—	—	10.12	19.72	2.02	3.44	6.08	35
AudioX-Base2026.06		1.56	—	—	—	—	—	—	9.48	19.12	2.06	3.17	6.21	31
AudioLDM-2-Large2026.06		2.05	—	—	—	—	—	—	16.32	13.86	2.1	2.95	6.35	30
AudioGen2026.06		2.48	—	—	—	—	—	—	15.94	11.09	2.16	3.3	5.45	29
MMAudio2026.06		2.5	—	—	—	—	—	—	11.52	17.83	2.17	3.02	6.12	32
Stable Audio Open2026.06		2.6	—	—	—	—	—	—	26	14.45	2.36	2.64	6.53	33
MAGNET-large2026.06		2.74	—	—	—	—	—	—	22.17	8.53	2.03	3.65	5.25	26
GenAU2026.01		3.44	33.03	36.76	32.9	33.6	27.04	21.1	34.55	10.1	2.52	—	—	—
Tango 22026.06		3.47	—	—	—	—	—	—	22.96	10	2.31	3.93	5.99	29
MMAudio2026.01		3.57	31.69	36.81	31.39	29.45	26.02	22.81	29.1	10.64	2.9	—	—	—
Siren + Plan-Critic2026.01		3.7	35.88	38.83	35.56	35.53	32.86	28.86	28.02	10.65	2.87	—	—	—
AudioX2026.01		4.44	33.93	37.71	33.6	32.32	29.61	28.52	35.86	9.39	2.94	—	—	—
Siren+BoN-162026.01		4.63	25.62	28.6	26.51	22.51	21.53	16.37	32.23	9.56	3.02	—	—	—
AudioLDM-L-Full2026.06		7.05	—	—	—	—	—	—	31.15	6.52	2.41	2.93	5.99	27