IAMCCS_BoardMaker

Before you generate a cinematic shot, someone has to decide the beats - where the camera starts, what each moment of the shot is doing, how one beat flows into the next. IAMCCS_BoardMaker is that planning step, made editable. Per the source it's a "planner assistant that exports Shotboard-importable board JSON without images": you describe the shot as a list of timed beats, and it turns that into structured JSON that the IAMCCS Shotboard system (and the Cine nodes) can consume.

The working data is board_data, a JSON blob the node's UI edits for you. Each row is a beat with: label, duration, local_prompt (what this beat shows), bridge (how the camera/cut carries into the next beat), camera, transition, force (the strength of the beat's visual push), and use_guide / use_prompt toggles. The default value ships with a three-beat example - opening, middle, final - so you can see the shape immediately.

Around that JSON sit the settings:

• global_prompt - the overall shot description that anchors every beat (identity, camera philosophy, continuity).
• duration_seconds, frame_rate, image_width, image_height - the timeline and canvas contract. Default 9s at 24fps, 768x432 (a 16:9-ish workable starting canvas).
• default_force (0.28) - the fallback force for rows that don't set their own.
• guide_policy - which rows contribute guide frames: every_checked_row, first_last, all, or none.
• board_name - labels the board so multiple boards stay distinguishable.

Outputs: board_json (the full board for the Shotboard importer), timeline_data (the timed view of the same thing), row_count, duration_seconds, and report. Wire board_json into a Shotboard or Cine planning node and you've gone from "vague idea" to "executable shot plan" without touching a paint tool.

Why you'd reach for it: if you're the kind of person who storyboards in text - beats, camera moves, transitions - this is the cleanest path into IAMCCS's video pipeline. It's deliberately image-free (the source calls that out), which is the point: the board is a plan, and the generation side makes the pictures. If you want pictures in the board, the DialogueFoley variant and the Shotboard editor are the next steps up.

Installing: part of IAMCCS/IAMCCS-nodes. ComfyUI Manager → search "IAMCCS", or cd ComfyUI/custom_nodes && git clone https://github.com/IAMCCS/IAMCCS-nodes.git, restart. No models, no extra deps - it's a structured editor over JSON.

The trap: this node produces a plan, not frames. New users connect board_json to a sampler and wonder why no video appears. Find the Shotboard/Cine node that ingests board_json and build the chain from the author's reference workflows; standalone, the board is just a well-formatted document. Also keep row durations roughly equal - a 0.5s beat next to a 4s beat makes the generation scheduling awkward downstream, even though the node will happily accept it.