Class-conditional joint audio-video generation on VGGSound (test)
138.8gFVDAVTok-cJAVG
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| AVTok-cJAVGGen Type=AR, #Param Tokenizer=208.4M, #Param Generator=632.4M2026.06 | 138.8 | 56.58 | 1.319 | 0.206 | |
| OviGen Type=FM, #Param Tokenizer=988.6M, #Param Generator=17.3B2026.06 | 972.65 | 129.02 | 0.814 | 0.172 | |
| JavisDiTGen Type=FM, #Param Tokenizer=448.7M, #Param Generator=8.9B2026.06 | 1,040.28 | 268.51 | 1.33 | 0.195 |