🎞️SVD Advanced

Stable Video Diffusion is old news at this point - Stability's 2023 image-to-video model, long since superseded by Wan 2.2 and the LTX line in the "just make video" rankings. But if you're still running SVD in ComfyUI, there are exactly three numbers that decide how a clip comes out: steps, motion strength, and seed. This node, 🎞️SVD Advanced, puts all three in one box so you don't go hunting for them across a sprawled-out graph.

It's part of ComfyUI-SVD-ZHO, the WIP pack by ZHO that recreates the old stablevideo.com web app as ComfyUI nodes. The whole point of the pack is to give you the same controls the website had, laid out the same way - and on that site, "advanced" was literally just these three sliders.

What it actually does

Nothing, and that's the point. The source is one line: return (steps, motion_strength, seed,). It's a pure pass-through - a grouping node. You set the three values once, and it hands them downstream as three separate outputs. It doesn't touch a model, doesn't call anything, runs instantly. That sounds pointless until you've built an SVD workflow with a KSampler, conditioning, and a video decode all competing for attention; keeping the three knobs that matter in one labeled box is genuinely nice ergonomics, and it matches the web UI the pack is cloning.

The inputs that matter

• steps - 25 to 60, default 30. Sampling steps. SVD is happy around 25–30; 60 is slow with diminishing returns.
• motion_strength - 1 to 255, default 127. This is the SVD motion bucket, on the same 0–255 scale the model's conditioning expects. Low values give you a barely-moving clip; high values give you real motion and, past a point, more artifacts. 127 is a sane place to start.
• seed - standard ComfyUI fare. Same seed, same graph, same video. Change it to explore variations.

Outputs

Three INTs, named steps, motion_strength, and seed. Wire motion_strength into the SVD conditioning's motion bucket, and feed steps and seed into your sampler. In the pack's text-to-video flow, a T2I Chooser picks your start frame first, then these values drive the video pass.

Install

The pack installs like any custom node:

cd ComfyUI/custom_nodes
git clone https://github.com/ZHO-ZHO-ZHO/ComfyUI-SVD-ZHO

Then restart ComfyUI. Or use ComfyUI Manager: search ComfyUI-SVD-ZHO and hit install. There's no requirements.txt and nothing to download - this node has zero dependencies, and the actual SVD model plus its conditioning nodes are built into ComfyUI core. Find it under the Zho模块组/🎞️SVD category.

Gotchas

The big one: this node only outputs numbers. If you wire it up and get no video, the problem is upstream - you're still missing the core SVD pieces (the ImageOnly checkpoint loader, the SVD conditioning node, the sampler). Also, don't plug these outputs into a modern Wan or LTX graph and expect magic; motion_strength means nothing to those models. And keep expectations in check: the pack is version 0.9 and explicitly WIP. For recreating that Stable Video 1.1 web workflow locally, though, this is the easiest way to get the same controls - you're just one small, honest helper node away.