Wave Forcing: Towards Speed-of-Light Streaming Video Generation

Yijie Jin; Yuxuan Liu; Jiajun Xu; Ye Huang; Chuzhi Wang; Jiayang Li; Ruiyang Chen; Chenling Hong

cs.CVOct 31, 2026

Wave Forcing: Towards Speed-of-Light Streaming Video Generation

Authors: Yijie Jin, Yuxuan Liu, Jiajun Xu, Ye Huang, Chuzhi Wang, Jiayang Li, Ruiyang Chen, Chenling Hong,

Organizations: DENG Lab, Shanghai Jiao Tong University · Xi’an Jiaotong University · Peking University · East China Normal University ·

Abstract

Real-time long-video generation requires continuous output with consistent appearance and natural motion. Existing chunk-wise autoregressive methods face challenges in both generation efficiency and long-video quality: serialized denoising and history-cache refresh require two forward passes even with one-step denoising, and longer rollouts can degrade quality and cause camera drift. We propose Wave Forcing (WF), which co-designs generation, training, and execution for continuous video output with multi-step denoising. WF uses chunk-wise causal attention to access earlier, cleaner chunks before their denoising is complete, and distributes staggered denoising and cache-refresh stages across full-model replicas to achieve 1F1C: one output chunk per concurrent forward round after pipeline fill. To adapt the model to this continuous generation process, we develop a three-stage training procedure that successively introduces mixed-noise denoising, motion supervision, and causal adaptation using self-generated history. To execute this process efficiently, Wave Runtime (WaveRT) uses Wave Parallelism (WP) for early K/V delivery and communication-computation overlap, together with optimized kernels and pipelined VAE decoding. Experiments on Wan2.1-T2V-1.3B show that 5-step WF maintains overall quality comparable to Rolling Forcing while achieving higher Dynamic Degree on both short and long videos. Our system efficiency experiments show that 4-step WF achieves 147.6 DiT FPS and 134.8 end-to-end FPS on eight H200 GPUs, yielding 7.07x the single-GPU end-to-end throughput. In DiT-only tests with matched SageAttention+W8A8 backends, denoising steps, and GPU counts, WP delivers up to 2.53x the throughput of the faster of the Ulysses and Ring context-parallel baselines across six model scales (1.3B-14B).

Figures & tables

Figure 1: Causal wavefront and continuous output (

S=4

). (a) Denoising and store advance on anti-diagonals, with dashed cells marking the next round. (b) One later-layer K/V edge arrives before use. Other edges and initial waits are omitted. (c) Both schedules retain four denoising forwards plus store per chunk. Over five rounds, one SF replica emits one chunk, whereas five WF replicas emit five after fill. Wall-clock gains depend on stage costs and overlap.

Panel of Figure 2 — Figure 2: Training flow. Stage 1 applies DMD to one sampled RF or SF rollout. Stage 2 re-noises teacher chunks and applies DMD and LPIPS to the same prediction $\widehat{Y}_{\theta}$ . Stage 3 samples one self-generated RF, WF, or SF rollout for on-policy DMD.

Figure 3: Hiding K/V delivery with identical schematic operation costs. (a) Layer barriers expose communication. (b) Early publication hides later waits, not dependencies or the initial wait. (c) Direct writes to receiver slots precede readiness signals and consumption.

Figure 4: Streaming VAE schedules. Panels (a,b) compare blocking and asynchronous decoding with matched costs. Panels (c,d) compare unbalanced and load-balanced PP under identical workloads and arrivals, with darker teal indicating longer stage service times. Panel (e) shows profile-guided contiguous decoder partitions. Each comparison pair uses its own schematic time scale, with transfers omitted and partition boundaries shown schematically. The first output incurs pipeline fill latency (Appendix B.5 ) .

			Standard VBench $\uparrow$				VBench-Long $\uparrow$		Stability		FPS $\uparrow$
Method	Steps	GPUs	Total	Quality	Semantic	Dynamic	Quality	Dynamic	Subj. $\uparrow$	Drift $\downarrow$	DiT	E2E
Rolling Forcing	5	1	81.15	83.77	70.66	51.39	81.18	36.63	92.75	0.17	29.12	17.46
+ Causal Attention	5	8	80.89	83.35	71.05	37.50	80.50	46.60	91.26	0.40	104.85	96.56
Causal Forcing (Full Attn.)	4	1	79.61	82.83	66.72	100.00	82.15	96.61	71.11	11.14	33.88	19.06
Wave Forcing (Ours)	4	8	80.74	83.93	67.95	84.72	80.89	74.97	91.88	0.36	147.58	134.81
+ SageAttention		8	81.03	84.02	69.07	65.28	81.16	77.24	90.79	0.36	147.78	134.79

Table 1: Short- and long-video comparison. Dynamic is Dynamic Degree. Subj./Drift are Subject Preservation/Camera Drift (scores: 0–100). The two FPS columns report DiT-only and end-to-end (E2E) runs.

Figure 5: Qualitative comparison. Rows show Causal Forcing, Rolling Forcing, and 4-step and 5-step Wave Forcing.

Model	Quality $\uparrow$	Dynamic $\uparrow$	Subj. $\uparrow$	Drift $\downarrow$
w/o off-policy DMD	76.68	15.73	76.53	0.26
w/o RF	80.19	88.82	77.48	6.24
w/o WF	80.34	48.12	95.64	0.10
w/o SF	75.35	23.66	84.04	1.89
Ours full	80.89	74.97	91.88	0.36

Table 2: 4-step training ablations on long videos. All rows evaluate final models. Quality/Dynamic are VBench-Long scores.

Figure 6: Qualitative training ablations. We compare the full 4-step model with each training ablation.

Figure 7: System efficiency with dummy DiT weights. (a) E2E throughput of 4-step WaveRT (8 GPUs) and the single-GPU baseline, with separate axes for 1.3B and 14B. (b) Steady-state tick p50 on the same WF graph, normalized to Sync. (c,d) BF16 DiT throughput across scales and WP’s percentage gain over the faster CP baseline. (e,f) DiT throughput across compute backends at 1.3B and 14B. Panels (c–f) disable VAE and use 5 steps/6 GPUs for 1.3B–9B and 4 steps/5 GPUs for 14B. Full measurements and repetition details appear in Tabs. 4 and 6 .

Appendix figures & tables9 assets

Supplementary material from the paper’s appendix.

Appendix

Symbol	Meaning
$N_{\mathrm{chunk}},i$	$N_{\mathrm{chunk}}$ is the number of requested chunks, and $i$ indexes chunks with $0\leq i<N_{\mathrm{chunk}}$ .
$S,s,\nu_{s}$	$S$ is the number of denoising updates, $s$ indexes denoising stages with $0\leq s<S$ , and $\nu_{s}$ is the diffusion timestep. Store occupies execution stage $s=S$ .
$k,r$	$k$ indexes forward rounds, and $r$ indexes diffusion ranks with $0\leq r\leq S$ , where rank $S$ is Store.
$Z_{i}^{(s)},B_{i}$	$Z_{i}^{(s)}$ is the state after $s$ updates, and $B_{i}=Z_{i}^{(S)}$ denotes the clean endpoint.
$\mathcal{A}_{k},\mathcal{F}_{k}$	active denoising-stage indices and their input states
$\mathcal{W}_{k}$	complete wavefront state, including the clean input to store when active

Appendix

Table 3: Graph and runtime notation. The complete wavefront includes store, whereas the denoising subset does not .

Configuration	GPUs	Tick p50	Steady	DiT	E2E	Speedup
		(ms)	(FPS)	(FPS)	(FPS)
1.3B, 4-step
Baseline	1	–	–	33.86	19.06	$1.00\times$
WaveRT (Sync)	8	91.96	130.50	129.31	127.06	$6.67\times$
WaveRT	8	78.88	152.13	147.58	134.81	$\pagecolor{tablebestbg}{\color[rgb]{0.2578,0.4844,0.4531}\mathbf{7.07\times}}$
WaveRT + Sage	8	78.36	153.15	147.78	134.79	$7.07\times$

Appendix

Table 4: Runtime throughput. WaveRT uses the WF graph and paged K/V unless marked Sync, while Baseline is single-GPU rolling inference. Both use BF16 with FA3 unless a kernel suffix is specified. GPU counts include denoising, Store, and VAE: 4+1+3 for 4-step and 5+1+2 for 5-step. Speedup is E2E relative to the same-scale, same-step baseline. All rows use dummy DiT weights and a real-weight VAE. Best values are determined within each model/step group using unrounded measurements. DiT/E2E are paired .

Backbone	Compute	Sync	Overlap	One-sided	Staggered	Paged
1.3B	BF16	91.956	85.730	81.669	81.953	78.882
1.3B	Sage	91.441	84.103	80.934	81.062	78.356
1.3B	Sage+W8A8	88.714	81.595	78.030	78.081	76.052
14B	BF16	464.270	449.689	425.907	425.188	413.795
14B	Sage	467.526	439.988	426.215	425.601	414.029
14B	Sage+W8A8	407.215	379.576	365.922	366.697	354.447

Appendix

Table 5: WaveRT same-graph system ablations report pooled steady tick p50 (ms, lower is better) for the 4-step, 8-GPU (4+1+3) setup across three fixed repeats. Each row compares execution strategies under a fixed compute backend. All runs use dummy DiT weights and a real-weight VAE and measure execution cost only .

Scale	Compute	Steps	GPUs	Ulysses CP	Ring CP	WP	WP/CP
				DiT throughput (FPS) $\uparrow$
1.3B	BF16	5	6	80.60	41.19	150.09 †	$1.862\times$ †
1.3B	Sage	5	6	84.01	56.99	151.15	$1.799\times$
1.3B	Sage+W8A8	5	6	61.17	45.96	154.75	$2.530\times$
3B	BF16	5	6	75.38	38.20	88.86	$1.179\times$
3B	Sage	5	6	75.93	37.00	89.26	$1.176\times$

Appendix

Table 6: DiT throughput across parallel strategies and compute backends, using dummy DiT weights with VAE disabled. Step counts and GPU counts are matched across methods within each row. FPS is the arithmetic mean of per-run rates, and best values are determined using unrounded measurements. WP/CP divides WP throughput by that of the faster CP baseline. Amber

{\color[rgb]{0.582,0.4219,0.207}\dagger}

marks supplemental measurements whose repetition and stopping protocols are described in Appendix D.2 .

Scale	Compute	Sync	Overlap	One-sided	Staggered	Paged
1.3B	BF16	91.946–92.009	85.659–85.763	81.589–81.744	81.842–82.107	78.851–78.967
1.3B	Sage	91.390–91.462	84.035–84.168	80.933–80.953	80.806–81.235	78.290–78.400
1.3B	Sage+W8A8	88.579–88.781	81.505–81.634	77.975–78.099	77.895–78.291	75.769–76.284
14B	BF16	463.785–464.761	449.001–450.082	425.517–426.670	424.276–425.844	413.094–414.374
14B	Sage	466.946–468.320	439.004–440.453	425.819–426.635	425.420–425.753	413.711–414.304
14B	Sage+W8A8	407.121–407.384	379.268–380.479	365.482–366.516	366.552–367.034	354.189–354.720

Appendix

Table 7: Minimum–maximum of per-run tick p50 (ms) across three fixed repetitions on the same node for causal 4-step, 8-GPU WaveRT with dummy DiT weights. These ranges are not confidence intervals.

Diagnostic	Measurement	Value
14B critical consumer	r0 forward interval	414.23 ms
14B producers	r1/r2/r3/store forward interval	370.50–379.37 ms
Aggregate rank-time difference	Critical minus mean of other ranks	38.32 ms
BF16 K/V communication	Rank 0 collective time summed over layers	58.45 ms
FP8 K/V communication	Payload and scale transfer, excluding conversion	31.23 ms
Copy-engine P2P	95,846,400-byte IPC copy	391.49 GB/s

Appendix

Table 8: Diagnostic-only component profiles of stage service, K/V transfer, receiver-side assembly, and VAE partitioning, using three-run means. Each row reports a component timing within its stated measurement scope. Because these components can overlap during execution, their durations cannot be summed to obtain E2E latency. The producer range spans rankwise means.

Schedule	Context state	Across-stage transfer	Execution structure
Self Forcing	Completed clean-history K/V	None	Sequential chunks
Live Avatar	Fixed-timestep stage-local K/V snapshots	Updated latent	Stage pipeline
HiAR	Refreshed output-timestep K/V within each hierarchy	Updated latent	Causal hierarchy pipeline
Rolling Forcing	Clean history and bidirectional live K/V	All active K/V if distributed	Joint mixed-noise denoising
Wave Forcing	Clean history and causal live K/V	Cleaner-to-noisier K/V and latent	Directed wavefront

Appendix

Table 9: Conditioning and exchanged state in different generation schedules. Live K/V is produced by active chunks at the current Transformer layer.

Method	Partitioned object	Main communication	Main constraint
Tensor parallelism	Hidden/head dimensions	Per-operator collectives	Small local GEMMs at high degree
Ulysses CP	Tokens to attention heads	Two All-to-All phases	Head divisibility and per-layer collectives
Ring Attention	K/V sequence blocks	Multi-round K/V circulation	Enough tile compute to hide P2P
Chunk-owned sync	Active query chunks	Layer K/V All-Gather	Global layer barrier
Wave Parallelism	Chunk–stage tasks	One-way K/V and latent	Full-model replicas and sufficient overlap slack

Appendix

Table 10: Resource-partitioning strategies, communication patterns, and constraints. TP and CP preserve the chosen graph, while WP schedules the fixed WF graph.

Parameters	Layers	Heads	Hidden dim.	FFN dim.
1.3B	30	12	1536	8960
14B	40	40	5120	13824
3B	32	18	2304	10752
5B	30	24	3072	14336
7B	29	30	3840	15360
9B	26	36	4608	18432

Appendix

Table 11: Model configurations for the parameter-scale comparison. The upper group contains the standard Wan2.1 architectures, and the lower group contains the synthetic architectures.

Streaming Video Generation

Streaming video generation creates video incrementally as frames or chunks arrive, enabling low-latency interactive synthesis and editing.

See all 101 papers

Ms. Forcing: Efficient Streaming Video Generation with Multi-Scale Patchification and Attention
Jul 23, 2026Zekun Li, Xiaoyan Cong, Hongyu Li +5Video Diffusion Models Long-Horizon Video Generation
1Brown University · 2Massachusetts Institute of Technology · 3Meta
Speculative Forcing for Efficient Autoregressive Video Generation
Oct 15, 2026Hyunwook Choi, Seungryong Kim, Dongyoon Han +1Efficient Video Generation Autoregressive Video Generation
KAIST AI · NAVER AI Lab
Accelerating Video Generation Inference with Sequential-Parallel 3D Positional Encoding Using a Global Time Index
Mar 2, 2026Chao Yuan, Pan LiDiffusion Model Inference Acceleration Autoregressive Video Generation

Jul 23, 2026J/K move · Enter open · S save

Ms. Forcing: Efficient Streaming Video Generation with Multi-Scale Patchification and Attention

Zekun Li, Xiaoyan Cong, Hongyu Li, Zhiyang Dou+4

1Brown University · 2Massachusetts Institute of Technology · 3Meta

Video Diffusion Models Long-Horizon Video Generation Streaming Video Generation Multi-Scale Attention Diffusion Model Inference Acceleration

Open paper AskPDF arXiv

Streaming video diffusion models have made substantial progress toward interactive and dynamic world simulation, but the nested autoregressive and denoising loops of conventional next-frame generation hinder real-time deployment. Recent rolling-window methods pipeline denoising across multiple consecutive frames at different noise levels, improving throughput and long-horizon stability. However, they tokenize every state at the same fine spatial granularity, leaving substantial noise-dependent redundancy in the joint denoising window. We propose Ms.Forcing, an efficient streaming video generation paradigm that adapts spatial granularity to each state's noise level. Its Multi-Scale Patchification (MSP) assigns coarser patches to noisier states, reducing the active-window token count by 45%, while Multi-Scale Self-Attention (MSSA) matches the density of visible non-sink keys and values to each query scale to further reduce attention cost. Because both schedules are fixed by window position, Ms.Forcing retains a static, hardware-friendly computation graph. We further introduce Homogeneous-Noise-Level DMD (H-DMD), which assembles each fake video from clean predictions sharing the same source noise level, thereby reducing the mismatch between DMD training sequences and inference-time rollouts. The multi-scale design helps offset the additional training cost of backpropagating through overlapping windows. We include both quantitative and qualitative experiments to show that Ms.Forcing reaches 22.84 FPS on a single H200 GPU, 39.6% faster than Rolling Forcing, while significantly improving VBench scores in both short video and long video generation setting.

Abstract

Streaming Video Generation

Explore similar work

Ms. Forcing: Efficient Streaming Video Generation with Multi-Scale Patchification and Attention

Speculative Forcing for Efficient Autoregressive Video Generation

Accelerating Video Generation Inference with Sequential-Parallel 3D Positional Encoding Using a Global Time Index