FlowLong: Inference-time Long Video Generation via Manifold-constrained Tweedie Matching

Hugging Face Daily Papers Papers
long-video-generation inference-time tweedie-matching sliding-windows temporal-consistency video-diffusion manifold-constrained

Summary

A novel inference-time method for long video generation using overlapping sliding windows with Tweedie matching and stochastic early-phase sampling to improve temporal consistency and visual quality without additional training.

Extending the generation horizon of video diffusion models to long sequences remains a long-standing and important challenge. Existing training-free approaches fall into two categories: extensions of bidirectional models, which are tightly coupled to specific architectures and suffer from quality degradation over long horizons, and autoregressive models, which accumulate drift errors due to exposure bias and tend to produce repetitive motion patterns. To address these issues, we propose a novel but simple inference-time approach for long video generation that is architecture-agnostic and requires no additional training. Our method generates long videos via overlapping sliding windows, where predicted clean samples from adjacent windows are blended via Tweedie matching to enforce both manifold constraint and temporal consistency across overlap regions. Stochastic early-phase sampling then synchronizes per-window trajectories by injecting fresh noise after each Tweedie matching correction in the high-noise phase, before transitioning to deterministic ODE sampling to preserve fine-grained visual fidelity. Applied to various video generation models, our method generates videos several times longer than the native window length while outperforming both training-free and autoregressive baselines in temporal consistency and visual quality, and further extends to audio-video joint generation and text-to-3DGS without any fine-tuning.
Original Article
View Cached Full Text

Cached at: 05/22/26, 06:35 AM

Paper page - FlowLong: Inference-time Long Video Generation via Manifold-constrained Tweedie Matching

Source: https://huggingface.co/papers/2605.20910

Abstract

A novel inference-time method for long video generation using overlapping sliding windows with Tweedie matching and stochastic early-phase sampling to improve temporal consistency and visual quality.

Extending the generation horizon ofvideo diffusion modelsto long sequences remains a long-standing and important challenge. Existing training-free approaches fall into two categories: extensions ofbidirectional models, which are tightly coupled to specific architectures and suffer from quality degradation over long horizons, andautoregressive models, which accumulate drift errors due toexposure biasand tend to produce repetitive motion patterns. To address these issues, we propose a novel but simple inference-time approach for longvideo generationthat is architecture-agnostic and requires no additional training. Our method generates long videos via overlappingsliding windows, where predicted clean samples from adjacent windows are blended viaTweedie matchingto enforce both manifold constraint andtemporal consistencyacross overlap regions.Stochastic early-phase samplingthen synchronizes per-window trajectories by injecting fresh noise after eachTweedie matchingcorrection in the high-noise phase, before transitioning todeterministic ODE samplingto preserve fine-grained visual fidelity. Applied to variousvideo generationmodels, our method generates videos several times longer than the native window length while outperforming both training-free and autoregressive baselines intemporal consistencyand visual quality, and further extends toaudio-video joint generationandtext-to-3DGSwithout any fine-tuning.

View arXiv pageView PDFProject pageGitHub2Add to collection

Get this paper in your agent:

hf papers read 2605\.20910

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2605.20910 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2605.20910 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2605.20910 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

Memento: Reconstruct to Remember for Consistent Long Video Generation

Hugging Face Daily Papers

Memento is a subject-reconstruction-guided framework that improves long-form video generation by preserving recurring subjects through memory-based reconstruction and dual-query mechanisms, achieving state-of-the-art performance in long-term subject consistency and cross-shot coherence.