Segment Planner Linked (shared inputs)

IAMCCS_SegmentPlannerLinked is the "don't repeat yourself" version of the IAMCCS_SegmentPlanner. In a long LTX-2 extension workflow you often end up with the planner duplicated all over the graph - one for the first segment, one per continuation, one for the last, maybe one inside each SuperNode. With the plain planner, that's a lot of widgets to keep in sync, and one stale overlap value three copies in means a seam you'll chase for an afternoon.

This variant forces its planning settings through inputs instead of letting them be hand-edited widgets. Feed it from a single shared source - usually the pack's IAMCCS_SegmentPlannerSettings node - and every copy of the planner agrees on segment_duration_s, planning_mode, segment_preset, and overlap_frames because they all read the same wires. Change the setting once and the whole graph re-plans.

How it works

Under the hood it's the exact same plan() math as the base planner - song duration → total frames, 8n+1 LTX rounding via ltx_round_mode, first/continuation raw frame split, per-index current-segment numbers. The only real difference is plumbing: segment_duration_s, planning_mode_in, segment_preset_in, overlap_frames_in are all forceInput, so they're link-only. There's also an optional ltx_round_mode_in if you want to push the rounding mode too; leave it unconnected and the widget value wins.

Inputs and outputs

Set song_duration_s and fps like normal, then link the four settings inputs from your shared SegmentPlannerSettings node (or anywhere - they're plain FLOAT/STRING/INT wires). segment_index drives which segment's numbers you get this iteration.

Outputs are identical to the base planner, and that's the point - drop it in place of a SegmentPlanner and nothing downstream changes:

• total_frames, estimated_segments, continuation_loops, first_segment_raw_frames, continuation_raw_frames, last_segment_unique_frames
• current-segment set: current_segment_raw_frames, current_segment_start_frames, current_segment_end_frames, current_segment_start_s, current_segment_end_s, current_remaining_frames_after, current_segment_report
• recommendation outputs: recommended_overlap_frames, recommended_audio_left_context_s, recommended_extension_preset

Wire current_segment_raw_frames into your sampler's length and the range nodes, and current_segment_start_s/_end_s into anything that needs to know where in the song you are.

Installing

Same pack as everything else: ComfyUI Manager → search "IAMCCS", or

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

then restart. No extra models or pip installs for this node - it's pure arithmetic. The LTX-2 model and VAEs that actually generate the video are the heavy part, and they're on you.

Gotchas

The biggest one is a wiring trap rather than a math trap: if you link planning_mode_in and segment_preset_in from a settings node but leave segment_duration_s on its widget, you can accidentally get preset-driven durations that contradict the duration you think you set. When in doubt, link all four from the settings node and let it own the values - that's the workflow this node was built for. And remember segment_index still needs to advance each loop iteration, or every segment gets the same start frame.