Using frame-to-frame differences as an additional CLIP condition improves video-to-audio generation quality and temporal alignment over raw-frame-only conditioning.
InICML, 62596–62626
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Visual Representation Matters: Exploiting Temporal Differences in Video-to-Audio Generation
Using frame-to-frame differences as an additional CLIP condition improves video-to-audio generation quality and temporal alignment over raw-frame-only conditioning.