- Who it helps
- Researchers and infrastructure teams studying continuous video generation.
- Availability
- October 31 paper and SJTU laboratory index are public. Deployment readiness and complete installation requirements were not verified.
What the reported speed means
The authors report 134.8 end-to-end frames per second with eight H200 GPUs in a four-step configuration. That hardware and configuration belong with the number; it is not a single-device performance claim.
WaveForcing pipelines video generation across multiple GPUs ↗
Quality and speed need separate checks
The paper also reports a five-step quality comparison using a small Wan model. Those quality results should not be silently assigned to the different four-step speed setup.
WaveForcing pipelines video generation across multiple GPUs ↗SJTU DENG laboratory project index ↗
What you can do
- Read the paper’s hardware and evaluation conditions. Open reference ↗
Where the evidence stops
Abstract and creator index read November 3. No weights downloaded, model run or speed measurement reproduced.
Sources read. Sources read November 3, 2026 UTC. No third-party software, model inference or full workflow executed.
Sources & further reading
Paper · paper
WaveForcing pipelines video generation across multiple GPUs ↗
Identifies the subject, documented capability or specific verification limit.
arxiv.org · page checked 2026-11-03
Hugging Face · official
SJTU DENG laboratory project index ↗
Identifies the subject, documented capability or specific verification limit.
SJTU DENG Lab · page checked 2026-11-03
Coverage note: Independent demonstrations and complete workflow verification remain unavailable.
The record over time (1)
Dated source checks and published explanations. An observation date can differ from the original release date.
- · Added to Current
WaveForcing overlaps video-generation work across several model copies to increase streaming throughput.