ResearchBenchmarksVideo-to-Speech Generation on CinePile OOD film dataVideo-to-Speech Generation on CinePile OOD film dataFollow58.7WERHiCoDiT57.51665.50873.581.492Apr 17, 2026Evaluation resultsMethodMethodLinksWERMCDDNSMOSEmotion Matching ScoreSpeaker Identity ScoreLSE-DHiCoDiT2026.0458.79.83.58250.17.6AlignDiT2026.0480.811.43.275.258.58.23EmoDubber2026.0488.39.92.876.545.17.72