IAMCCS Cine Shotboard Planner Pro

The standard LTX-2.3 workflow has a real interface problem: the model conditions on a time-windowed first-frame/last-frame basis, but stock ComfyUI has no clean way to say "this frame at second 2, this reference at second 5, this camera move at second 7." IAMCCS Shotboard Planner Pro is the pack's answer, and it's the one you'll see dropped into most IAMCCS cine graphs. It's a planner-only node: it takes your shot table and emits a complete plan - no conditioning, no latent, just the plan, wrapped in a single cine_linx output that a backend can execute.

That planner-only design is deliberate. The original Timeline Pro variant did everything in one node, which works until you also need to feed its output back upstream - then you get a graph cycle. Planner Pro sits upstream of PromptRelay precisely so you don't. The docstring says it plainly: "Planner-only Shotboard Pro. Use upstream of PromptRelay to avoid graph cycles."

How it works. timeline_data holds the rows the UI edits: time, image reference, force, label, camera move, transition, note and a guide-enable flag. From those rows the node builds guide frames and per-segment local prompts, then bundles everything into cine_linx (the pack's LINX contract dict) along with the settings below.

The inputs that actually matter:

• guide_policy - safe_core_guides (default) is the conservative pick: only a few well-spaced guides fire, which keeps things stable. every_checked_row fires every checked row, and prompt_only skips image guides entirely and relies on the global prompt. If you're fighting drift, start with safe_core_guides and raise max_guides (default 5) only when you need more choreography.
• min_guide_gap_seconds - default 1.75s. Forces a minimum spacing between guides; set lower if you're doing fast cuts.
• default_force - the guide strength when a row doesn't set its own. 0.22 is a good conservative default; the range is 0–1.
• promptrelay_epsilon - tolerance for the PromptRelay segment timing (0.001 default). This is the value that decides when a "this segment's prompt now" switch fires.
• ltx_round_mode - up_8n_plus_1 (default) rounds your frame count up to LTX's 8n+1 requirement. nearest_8n_plus_1 and none are there if you'd rather round to nearest or handle the math yourself.

The image inputs (image_paths, image_width, image_height, image_resize_method, image_multiple_of, img_compression) control how your reference stills get normalized before they're baked into the plan. image_resize_method defaults to crop, and image_multiple_of defaults to 32 to stay LTX-friendly. Feed an image batch through the optional multi_input instead of paths if you prefer.

Output: the single cine_linx socket. Wire it into a backend node (the V4 backend, or the Cine FLF backend path) - or into CineInfo if you want to see the report of what it decided.

Install and gotchas. Same as the whole pack: ComfyUI Manager (search "IAMCCS"), or git clone https://github.com/IAMCCS/IAMCCS-nodes into custom_nodes, then restart. No bundled models - this is planning logic, and it expects your LTX-2.3 stack (model + video VAE + audio VAE) to be handled downstream. Keep the pack updated; the author releases frequently, and a stale install is the usual cause of a node that loads but validates wrong. Remember the 8n+1 tail: whatever duration you set, the generated clip will come back a few frames longer until the backend trims it.