== NODE TO GENERATE PROMPT FOR ==
location_id: {location_id}
location_description (English visual): {location_description}

node_id: {node_id}
kind: {node_kind}
label: {node_label}
description (English): {node_description}
shared_visual_anchors_with_parent: {shared_anchors_json}

== PARENT NODE ==
{parent_block}

== SHOTS COVERED BY THIS NODE ({shot_count}) ==
{shots_block}

== TASK ==
Write ONE photorealistic background image prompt for this node.
- Output strict JSON: {{"t2i_prompt": "<English string>"}}
- ENGLISH ONLY — no Korean, no Hanja, no kana, no proper names.
- NO people, NO blood, NO weapons, NO action — empty/quiet space only.
- Open the prompt as instructed in the system message (root vs. non-root).
- For non-root: explicitly mention each shared_visual_anchors_with_parent entry.
- For root: thoroughly describe wall finish, floor, ceiling, lighting, palette.
- Be concrete with frame elements (doors, windows, fixtures, furniture).
- One single instant — no collage.