FlowLong: Training-free Long Audio-Video Generation via Manifold-constrained Tweedie Matching
FlowLong is presented, a training-free, architecture-agnostic, inference-time framework that harmonizes overlapping windows sampled in parallel via Tweedie matching, a manifold-constrained blend of predicted clean samples on overlaps, and stochastic early-phase sampling that yields arbitrarily long, phase-locked audio-...
Jangho Park, G. Park, Gihyun Kwon et al.
· 0 citations