Wan Long Planner

Video models don't generate a minute of footage in one pass - you chunk it, chain it, and hope the seams disappear. IAMCCS_WanLongPlanner ("Wan Long Planner") is the node that does the planning arithmetic for you. You tell it how long the final video should be (total_duration_s), your fps, and how many visible frames each segment should use, and it answers the questions every long-video workflow needs: how many segments is that, which one am I on right now, is it the first or last, and what continuity settings should apply here?

It's a calculator with a really well-thought-out output contract. All the outputs are plain INTs/FLOATs, so it plugs into whatever nodes you're already using - it doesn't try to be a magic all-in-one.

The inputs you set

• total_duration_s (default 12), fps (default 16) - your target runtime and frame rate. The whole plan derives from these.
• first_visible_frames / continuation_visible_frames (both default 81) - the first segment is often longer (81 frames is the WAN convention for a "full" clip), and continuations carry a motion tail, so they get their own budget.
• segment_index - your loop counter. The node uses it to figure out where you are.
• first_motion_count / continuation_motion_count - how many prev-tail slots count as motion on the first segment vs. later ones.
• end_overshoot_slots (default 2) - extra slots rendered past the visible end so FLF end frames converge; feeds trim_slots.
• use_end_frame_mode - all, first_only, continuations_only, last_only, or off - decides which segments should bother locking an end frame at all.

The outputs that actually matter

• total_frames, estimated_segments, continuation_generations - the plan itself.
• is_first_segment, is_last_segment - 1/0 flags that drive routing (e.g., bypass prev_samples on the first segment).
• current_start_frame / current_end_frame / current_visible_frames - where this chunk sits in the final timeline.
• current_motion_count, use_prev_samples, use_end_frame, active_prompt_slot, current_trim_slots - the per-segment settings you feed straight into the motion nodes (WanImageMotionPro) and the prompt planner. That's the loop's decision-making, all pre-computed for you.

Install

Standard IAMCCS-nodes:

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

or ComfyUI Manager → search "IAMCCS" → restart. No models, no extra deps. ComfyUI ≥ 0.3.0, Python ≥ 3.12, Torch ≥ 2.8.

Where it pays off

If you've ever manually calculated "how many 81-frame segments is 40 seconds at 16fps" in your head mid-workflow, this node ends that era. It also removes the classic first-segment bug: the flags make it trivial to render the opening chunk without prev_samples and every later chunk with it. Just remember the math assumes your sampler actually produces the number of frames you budget - if your length widget doesn't match continuation_visible_frames, the plan is lying to you, so keep them in sync.