Native Wan Pose Strength

If you've done any pose-driven WAN work, you already know the tension: crank pose adherence up and motion follows your reference exactly but everything else - prompt, style, scene composition - starts fighting for control against a conditioning signal that's dominating the generation. Turn it down and the model listens to your prompt again but starts drifting off the reference motion. This node exists to give you that dial explicitly, on the model itself, rather than leaving it baked into whatever defaults the base pose-conditioning path shipped with.

What it's for

The "Native" in the name is doing real work - this operates on ComfyUI's own built-in WAN pose-conditioning path, not a VACE-based pipeline. If your workflow is driving WAN with pose control through VACE conditioning, strength there is set on the conditioning side, not through this node. This one wraps the loaded WAN MODEL itself and returns a patched version with pose-strength baked in, so it slots in between your model loader and your sampler.

How it works

It's a model patch: you feed in your WAN MODEL, set how strongly the native pose signal should influence generation, and get back a MODEL wired the same way but with that strength applied internally. The pack doesn't document the exact math beyond the parameter names and ranges, so take the following as a practical read rather than a verified algorithm description - but it's a reliable one to work from.

The inputs and outputs that matter

• model - your loaded WAN model, before the sampler.
• pose_strength (0–1, default 1) - the standard strength dial. At 1, pose conditioning applies at its normal, full weight. Lower it if pose is overpowering your prompt or producing stiff, over-literal motion; this is the first thing to touch if your outputs feel like they're ignoring everything except the reference pose.
• overdrive (0–1, default 0) - a second dial that goes beyond what pose_strength's normal 0–1 range covers. The name says the job: push it up when pose_strength at 1.0 still isn't giving you tight enough adherence to the driving motion and you need more than "full strength" gets you. Expect diminishing returns and a higher chance of artifacts or jitter the further you push it - this isn't a free lunch, it's trading stability for adherence.

Output: model, patched, ready to feed straight into your sampler.

How to install it

Search "Link Comfy Nodes" in ComfyUI Manager and install from there. Or by hand:

cd ComfyUI/custom_nodes
git clone https://github.com/Mister-Link/link-comfy-nodes
pip install -r link-comfy-nodes/requirements.txt

Restart ComfyUI. No extra model downloads for this specific node - it's a lightweight wrapper around whatever WAN model you've already loaded.

Common issues & troubleshooting

Wiring it into the wrong pipeline. If your pose control is coming through VACE conditioning nodes rather than WAN's native pose path, this node has nothing to grab onto - double-check which pose-driven approach your workflow actually uses before assuming this is the strength control you need.

Starting from an extreme value. With no tooltip documenting the exact behavior, start at the defaults (pose_strength 1, overdrive 0) and nudge one at a time rather than guessing at a combination. Small moves on overdrive in particular - it's easy to overshoot into visible artifacts before you've found the useful range.

Forgetting it has to sit before the sampler. Like any model-patching node in ComfyUI, this only affects generation if it's wired between your model loader and your sampler node - connecting it anywhere else silently does nothing.