ResearchBenchmarksText-to-Speech on Spark-TTS generated audioText-to-Speech on Spark-TTS generated audioFollow0.2057FAD (VGGish)None0.1999040.2390270.278150.317273May 25, 2026Evaluation resultsMethodMethodLinksFAD (VGGish)FAD (CLAP)MOS (NISQA)MOS (DNSMOS)WERCERNoneWatermarking=NoneWatermarking=None2026.050.20570.04013.372.940.970.18BaseWatermarking=BaseWatermarking=Base2026.050.22210.04843.32.970.980.12OursWatermarking=OursWatermarking=Ours2026.050.35060.04723.462.960.990.24