Nodes/ComfyUI-WanVideoWrapper/WanVideo ATI Tracks Visualize
ComfyUI Node Runs on cloud

WanVideo ATI Tracks Visualize

See your trajectories before you render

By kijai·Created about a year ago·Updated 2 months ago· 6,651
WanVideo ATI Tracks Visualize
  • images
  • images
tracks
min_radius1
max_radius6
max_retain50

Trajectory strings are not human-readable. A tracks value that's meant to describe a hand sweeping across frame is, in its raw form, just numbers - and debugging motion control by staring at coordinates is a bad time. This node exists to make that debugging visual: give it your images and your tracks, and it draws the trajectory points directly onto the frames as dots, so you can see at a glance whether the path you meant to draw is the path that's actually going to condition the generation.

What it's for

This is the companion visualization node for WanVideo ATI Comfy, the trajectory-based motion patcher elsewhere in this pack. The workflow it's built for: build or generate your tracks string, run it through this node against your source frames before sending it into ATI's model patch, look at the overlay, and only then commit to a full generation. Catching a garbled or off-frame trajectory here costs nothing; catching it after a multi-minute Wan render costs real time and, if you're paying for GPU by the second, real money.

The inputs that matter

images (IMAGE, required) - the frames you're checking the trajectory against; typically your source or reference clip, at the same resolution you'll actually be generating at, since track coordinates are resolution-dependent. tracks (STRING, required) - the same trajectory data you're feeding into ATI. This is a plain string format specific to the ATI system, not the TRACKS object type used by this pack's separate WanMove motion nodes - the two aren't interchangeable, so make sure you're visualizing the right kind of track data for the system you're actually using downstream.

Three drawing parameters, all cosmetic rather than functional: min_radius and max_radius (defaults 1 and 6) set how small and large the drawn track markers get, and max_retain (default 50) controls how many trailing points stick around in the visualization - useful for seeing the shape of a path over time rather than just a single dot per frame. None of these three affect the actual generation; they only change how legible the overlay is to you.

Output is a single images - your frames with the trajectory dots drawn on, ready for a Preview Image node.

How to install it

Through ComfyUI Manager, search WanVideoWrapper and install; manually:

cd ComfyUI/custom_nodes
git clone https://github.com/kijai/ComfyUI-WanVideoWrapper

then restart. Nothing else to fetch - this node is pure visualization and doesn't touch a model or checkpoint.

Common issues & troubleshooting

Tracks visualize in the wrong place relative to the frame. Since track coordinates are resolution-dependent, check that the images you're feeding this node are at the same resolution as your intended generation. A trajectory built for one size and previewed against frames of a different size will look wrong here even if it would have worked correctly at the matching resolution.

Nothing draws at all. That almost always means the tracks string is empty, malformed, or was built for the wrong track format (WanMove's TRACKS object rather than ATI's plain-string format). Trace back to whatever produced the string before assuming this node is broken.

It's a preview, not a validator. This node won't tell you whether a trajectory will produce good-looking motion - only whether the path is where you think it is. Motion quality still depends on the model, the temperature/topk settings on WanVideo ATI Comfy, and how well the trajectory itself matches physically plausible movement.

CategoryWanVideoWrapper

Inputs (5)

NameTypeDefaultDescription
imagesIMAGE
tracksSTRING
min_radiusINT10–100radius for the very first point (oldest)
max_radiusINT60–100radius for the current point (newest)
max_retainINT500–100Maximum number of points to retain

Outputs (1)

NameTypeDescription
imagesIMAGE