IAMCCS Cine Shot Line Builder

If the shotboard planners are for single continuous shots, Shot Line Builder is for the other approach: generating each shot as its own piece and assembling a scene from pieces. This is the node that writes one shot's worth of direction - framing, subject, action, dialogue, camera, lighting - and turns it into a structured "shot line" that a multi-generation director can stack with its siblings. The defaults even read like a scene: "Mara," "medium close-up," "near the kitchen window, she speaks softly but firmly," "You left the city without saying goodbye." The author clearly built this while writing dialogue scenes.

How it works. You describe the shot through a row of discrete fields, and the node composes them into two outputs: shot_line (the machine-readable line for the multigen pipeline) and shot_prompt (the human-readable prompt that expresses the whole shot). The reference_token output identifies which reference image the shot should use, and report dumps the composition for debugging.

The inputs that matter. The full list is long, but they're all fields you actually understand:

• cut_mode - hard_cut, continuity_cut, soft_cut, or match_cut. How this shot joins the last one.
• framing - a ten-way dropdown from "wide establishing shot" down to "insert detail" and "two-shot," plus custom.
• subject and action_prompt - who's in the frame and what they're doing.
• dialogue_text and voice_direction - what they say and how (this feeds the audio/lipsync side of LTX-2.3).
• camera_movement and lighting - dropdowns with a custom escape hatch each.
• reference_mode and the primary_reference_index/secondary_reference_index - which references the shot anchors on (single, start_to_end, or dual).
• audio_index - which audio line the shot is synced to.
• continuity_note - free-form consistency reminders ("keep coherent eyelines and stable character identity").
• duration_seconds - defaults to 4, the classic shot length.

Two optional overrides: prompt_override (per its tooltip, if connected or filled it replaces the dropdown-composed shot prompt entirely) and extra_prompt (appends extra text).

Where it fits. Build shots with this node, then feed their shot_line outputs into the pack's multigen director/line-stacker nodes (the IAMCCS_CineLineStacker / multi-generation side of the family) to sequence a whole scene. It's the "write one shot" atom of the multi-generation workflow, the same way the planners are the "write one continuous shot" atom.

Install. Pack-wide: ComfyUI Manager (search "IAMCCS"), or git clone https://github.com/IAMCCS/IAMCCS-nodes into custom_nodes, restart. Pure prompt-composition logic - no models, no VRAM, no LTX stack required just to use it.

Gotchas. The defaults are a temptation: the dropdowns have custom options, and the prompt_override tooltip is explicit that filling it bypasses the whole composition. If your shots keep coming out generic, you're probably leaning on the defaults instead of writing the action_prompt/dialogue_text/continuity_note fields. And remember the prompt-sensitivity reality of LTX-2.3: it rewards specific, detailed prompts, so a half-empty shot line will produce a half-committed shot. Fill in the fields.