ComfyUI-Ref2VA-VSA is an experimental custom node that applies video-sparse attention to MiniMax H3's reference-to-video path. Its useful promise is a specific local character-video recipe: keep multimodal reference conditioning separate while pruning selected attention work in generated video. It is not a new hosted H3 model or a guarantee that every character will remain unchanged.
The author's repository supplies a workflow, gate extraction utility, file list and benchmark. This guide reviews those materials as of October 20, 2026. We have not run the workflow. H3Vid's online Ref2VA tool has a separate execution path and does not use this node.
What the 72-second number means
The README's warm result is approximately 72 seconds for 124 frames at 24fps and 1344×768 on an RTX 4090 with 24GB VRAM. Its first run is about 95 seconds. A 124-frame output is approximately 5.17 seconds, which explains the rounded five-second label.
The listed peak memory is about 13.5GB, but the requirements still specify a 24GB NVIDIA GPU. Peak memory from one measured configuration does not validate all 16GB cards, different text encoders, larger reference sets or other installations. A social repost's broader sixteen-gigabyte claim should not replace the repository's actual tested machine and requirements.
The comparison also changes more than attention: the VSA case uses four Euler/simple steps, while the VDN comparison uses eight steps with another sampler and scheduler. Read it as a comparison of complete author-selected recipes, rather than a controlled measurement of one kernel change. Ratios based on first-run timing also differ from ratios based on warm runs.
Why reference conditioning needs special handling
Reference-to-video packs text and reference information alongside the sequence being generated. Applying video tiling indiscriminately can mix conditioning segments into a geometry intended for video. The project separates the conditioning prefix from generated-video tiles and keeps that prefix outside the pruning operation.
This is an architectural explanation of the node's intent. It does not establish perfect identity preservation. Faces, clothes, pose and interactions still need visual review in the resulting clip. A reference remaining available to attention is different from the generator obeying it exactly in every frame.
The patch transplants learned gate projections from a FastH3 VSA checkpoint onto the Ref2VA model. That gate file is a distinct component from the Turbo LoRA: one controls sparse selection, while the other supplies the accelerated sampling recipe. Renaming an ordinary LoRA to the gate filename will not create those projections.
Assemble the required files
Use the current repository's linked files and destinations. The documented recipe includes a pruned INT8 Ref2VA diffusion model, a four-step Ref2V Turbo LoRA, fasth3_vsa_gate.safetensors, an NVFP4 AWQ Qwen3-VL text encoder, FP16 video VAE and FP32 audio VAE. The diffusion model belongs in models/diffusion_models; gate and Turbo files go under models/loras; encoder and VAEs have their own directories.
The complete filenames and download links remain in the author's requirements section. Check every selector in the workflow against the actual file on disk. Similar H3 names can refer to a different task, quantization or conditioning contract.
The gate extraction utility accepts the named FastH3 VSA source checkpoint and writes the small gate file. Use the script's documented source rather than assuming every checkpoint contains suitable gate tensors. Keep a record of which source produced it so the experimental setup can be reconstructed later.
Install and preserve node order
Install the custom node inside ComfyUI's custom_nodes, restart ComfyUI, then load workflows/ref2va_vsa_4step_rtx4090.json. The README also requires a current native-H3 ComfyUI installation, PyTorch/CUDA and comfy-kitchen with CUDA sol_attn support. A visible node in the menu does not prove that the required compiled attention kernel is available.
The intended model path is: load the Ref2VA INT8 model, apply the four-step Turbo LoRA at strength 1.0, apply Ref2VAVSAGatePatch at sparsity 0.75, then apply H3 sigma shifts. The documented shifts are 12 for video and 3 for audio. Sampling uses four steps, the simple scheduler and Euler sampler.
For an initial check, use the supplied example image and saved workflow before substituting your own references. Preserve its wiring and change one input at a time. If the node cannot import or the kernel fails, capture the software versions and error; changing sparsity will not repair a missing binary dependency.
Evaluate a character shot deliberately
After the baseline works, try a short original scene with one adult character and a simple action, such as turning toward a window and placing a cup on a table. Use a clear reference that shows the face and the clothing details you intend to retain. Keep the prompt consistent with that reference rather than asking it to anchor incompatible identities.
Save the workflow, seed, files and full timing log with the output. Review the beginning, middle and end for face shape, accessories, hand-object contact and audio continuity. Then watch the entire clip, because a few attractive frames cannot reveal every temporal defect.
For a fair local speed comparison, separate first-use initialization from repeated warm runs and keep references, frame count and canvas fixed. Do not add several accelerators together on the first attempt: a working baseline is what makes a later regression understandable.
This experimental gate transplant addresses a different setup question from Veda attention or the broad ComfyUI installation guide. Choose the guide that matches the component you are actually installing, and use the original repository as the authority when its workflow changes.