ResearchBenchmarksMany-to-one Audio Generation ((T+I) → A) on Triplet text, image, audioMany-to-one Audio Generation ((T+I) → A) on Triplet text, image, audioFollow28.13CLAP Score (T->A)FlowBind3.918810.204416.4922.7756Dec 17, 2025Evaluation resultsMethodMethodLinksCLAP Score (T->A)AIS Score (I->A)FlowBind2025.1228.1376.02OmniFlow2025.127.6859.32CoDi2025.124.8561.28