Motion Bridge Load ????

Long-form WAN 2.2 video is chunked by necessity - native context stops around 81 frames, so anything longer means generating segments and joining them. The hard part was always the seam: the next segment needs the previous one's last frame (and a bit of motion context) to continue cleanly, and hand-carrying those latents across workflow runs was fragile. IAMCCS_MotionBridgeLoad is the "resume" half of the pack's answer: load a motion bridge saved to disk and get back the latent and junction frame you need to start the next segment.

It's the load-side sibling of IAMCCS_MotionBridgeSave, from the WAN section of IAMCCS-nodes - the pack by IAMCCS whose SVI Pro v2 work brought first-frame/last-frame control to the WAN scene in early 2026. If you're building chained long-form WAN, this pair is the seam management.

How it works

IAMCCS_MotionBridgeSave (in the previous segment's workflow) wrote a .safetensors to output/motion_bridges/ containing the tail latent slices plus the last frame. This node reads that file back and hands you three things:

• latent - the saved tail latents, ready to become your anchor_samples for the next segment.
• image - the saved last_frame, the junction image for conditioning or inpainting.
• tail_frames (INT) - how many frames were sliced, so your next workflow's motion_latent_count can match.

The bridge_name dropdown lists bridges from output/motion_bridges/ - and yes, you need to refresh the ComfyUI page after saving a new one, because the file list is scanned at page load.

Inputs and outputs

• bridge_name (enum) - the bridge file to load.
• image_override (IMAGE, optional) - the one input you'll actually reach for. If you want to inpaint or manually edit the junction frame before it becomes the next segment's anchor, connect your edited image here and it replaces the saved last_frame. The tooltip is explicit about the workflow: edit the junction frame, then feed it through VAEEncode into anchor_samples.

Outputs: latent (LATENT), image (IMAGE), tail_frames (INT).

How to install

Comes with IAMCCS-nodes. ComfyUI Manager search "IAMCCS", or:

cd ComfyUI/custom_nodes
git clone https://github.com/IAMCCS/IAMCCS-nodes.git

Restart. No model downloads; you need WAN 2.2 (or a WAN derivative) from elsewhere. Pack floors: ComfyUI ≥ 0.3.0, Python ≥ 3.12, Torch ≥ 2.8.

Gotchas

The most common stumble is the file list: save a bridge in one workflow, switch workflows, and the load dropdown still shows (none) until you refresh the page. That's documented in the tooltip, but it catches everyone once. Also, a bridge is a fixed hand-off - if you change the previous segment's tail_frames after saving, the numbers in the loaded bridge and your next workflow won't agree, and WAN's motion continuity will quietly drift. Match motion_latent_count on the next workflow's motion node to the bridge's tail_frames, and you're fine. If you'd rather not store the bridge on disk at all, the sibling MotionBridgeSave is the one that writes it - this node just reads.