Extensions/StrideAR
ComfyUI Extension

StrideAR

StrideAR: Training-free 2.04x-2.81x speedup on Krea Realtime 14B + Self-Forcing 1.3B (up to 23.1 FPS real-time on B200, 2.67x on H100) with lazy target KV caching and low-VRAM offload.

By statite-ai·Created 13 days ago·Updated 2 days ago· 0
statite-ai/ComfyUI-StrideAR
Nodes—
On cloudLocal install
Stars0
Updated2 days ago
Readme

ComfyUI-StrideAR

Runs StrideAR text-to-video on Krea Realtime 14B in ComfyUI. It adds one node, StrideAR Generate, and calls the installed stridear package.

Project page: statite.ai/stridear · Paper: PDF · Earlier preprint: PDF · Code: statite-ai/StrideAR

Install

git clone https://github.com/comfyanonymous/ComfyUI && cd ComfyUI
python3.11 -m venv .venv-krea && source .venv-krea/bin/activate
pip install -r requirements.txt "stridear[krea] @ git+https://github.com/statite-ai/StrideAR"
git clone https://github.com/statite-ai/ComfyUI-StrideAR custom_nodes/ComfyUI-StrideAR
stridear download --backbone krea --weights models/stridear/krea
python main.py

stridear pins exact library versions, including torch, so use a separate environment. The download is about 50 GB.

Use

  1. Open http://127.0.0.1:8188. If ComfyUI runs on a remote GPU machine, first run ssh -L 8188:127.0.0.1:8188 user@your-gpu-machine on your own computer.
  2. Go to Templates, find ComfyUI-StrideAR and open StrideAR. You can also drag example_workflows/StrideAR.json onto the page.
  3. Type your prompt in StrideAR Generate.

The StrideAR workflow in ComfyUI

  1. Click Run. The first run loads the models, which takes about 3 minutes. Later runs start right away.
  2. The video plays in Save Video and is saved in output/video.

A finished StrideAR video in ComfyUI

Settings

| Input | Meaning | |---|---| | prompt | The text prompt | | k (quality) | How much of the video the large target model generates; the small drafter generates the rest. Higher is slower, and 2 is the fastest setting | | seconds | Video length, rounded to the nearest 0.75 s. 26.6 s (426 frames) is the tested length | | seed | Random seed. Set it to fixed to reproduce a video | | weights | A folder under models/stridear (default krea). Symlinked folders are not listed; to keep weights on another disk, add a stridear path in extra_model_paths.yaml. Refresh the page to see a new folder | | keep_loaded | Keep the models loaded for the next run. Changing the prompt then takes under a second | | low_vram_offload | Keep only the active model on the GPU. On by default on GPUs under 75 GiB |

GPU memory

Measured on one H100 for a 26.6 s video:

| Setting | Peak GPU memory | Generation time, k = 2 | Generation time, k = 7 | |---|---|---|---| | low_vram_offload off | 44.5 GiB | 30 s | 39 s | | low_vram_offload on | 41.9 GiB | 34 s | 46 s |

Offload also ran within a 48 GB limit set on the H100. Either setting needs more than 40 GB of GPU memory, because the target model and its context cache must be on the GPU together. Offload keeps the model weights in system RAM, which we estimate at about 40 GB.

Tested on H100 (FlashAttention 3) and B200 (FlashAttention 4), which are selected automatically. Peak memory is the same on both, and on a B200 a 26.6 s video at k = 2 generates in 19 s. Other NVIDIA GPUs with enough memory need FlashAttention 2 (pip install flash-attn) and are untested. A different attention kernel can give a different video for the same seed, so the reference hashes apply only to FlashAttention 3 on Hopper GPUs.

Reproducibility

With FlashAttention 3 on an H100, the latents and 8-bit frames from this node matched stridear generate bit for bit, and the latents matched the stridear-benchmark reference hashes, with offload on or off. Save Video uses its own encoder, so the mp4 file differs from the one stridear generate writes, and encoding adds about 30 s at 426 frames.

Citation

@misc{hassan2026stridear,
  title  = {StrideAR: Strided Target Blocks for Fast Autoregressive Video Generation},
  author = {Hassan, Hudeifa},
  year   = {2026},
  note   = {Preprint},
  url    = {https://statite.ai/stridear/}
}

License

This node is Apache-2.0. It imports stridear and contains none of its code. StrideAR's Krea backend is CC BY-NC-SA 4.0, which allows non-commercial use only, and model weights carry their own licenses. See the StrideAR license notes.