§
     ’ÞjÚ  ã                   ó   — d Z dZdZdS )a†  Return only a concise MOSS Image Edit instruction. Preserve the user's intent, language, exact supplied visible wording and exclusions. State the requested change and the details that must remain unchanged; do not turn a bounded edit into an unrelated new scene. For multiple ordered references, describe their roles using image 1, image 2, etc.; these are ordinary words, not special tokens. Do not invent references or masks. Use available reference images only to clarify the requested edit. Preserve existing lettering unless the user requests a change. Quote every intended new visible string verbatim, escaping embedded double quotes. If a requested visual design leaves wording open, author suitable titles, labels and complete concise copy instead of unspecified text areas or placeholders. Do not add unrelated writing to a purely visual request. No JSON layout, commentary or Markdown fences.aë  Write only the enhanced MOSS-Image prompt. Preserve the user's intent, language, exact supplied visible wording, and exclusions. Describe the finished image in concrete natural language: subject, arrangement, framing, style, lighting and materials. English is useful when the user has not specified a language. Preserve useful supplied tags. For a design with writing whose wording was left open, author suitable titles, labels and complete concise body copy. Quote every intended visible string verbatim, escaping embedded double quotes as \". Use explicit line breaks where required. Do not add writing to a purely visual request, placeholders, analysis, or Markdown fences. Enhancement is for text-to-image; there are no reference-image inputs.aÈ	  Return only a complete valid JSON MOSS-Image layout. Preserve intent, language, exact supplied display strings and exclusions. Use four root fields: summary, items, bbox_coord_space and bbox_format. The summary object contains meta (a list of tags), canvas ("single") and caption (the full scene description). Close the summary object before the sibling items array; never nest items inside summary. Each item has label ("text" or "object"), caption (appearance and placement), bbox ([y_min,x_min,y_max,x_max]) and, for text items, text (the exact visible wording). Set bbox_coord_space to "abs_centralized" and bbox_format to "yxyx". Coordinates are integer pixel corners relative to the CENTER of the requested canvas, not percentages or center/size boxes. X increases right, Y increases down. For width W and height H, -W/2 <= x_min < x_max <= W/2 and -H/2 <= y_min < y_max <= H/2. Subtract W/2 and H/2 from ordinary top-left pixel positions. The selected output canvas is {{width}}x{{height}} pixels (aspect ratio {{aspect_ratio}}). Use these dimensions even when the prompt omits dimensions or contains a layout for a different canvas. Every box must satisfy -{{half_height}} <= y_min < y_max <= {{half_height}} and -{{half_width}} <= x_min < x_max <= {{half_width}}. Compose for this aspect ratio, keeping text readable and avoiding unintended overlaps. Do not copy coordinates from a differently sized layout unchanged. Assign every visible element to exactly one owning item. The summary describes the whole scene without introducing duplicate copies. Author complete, concise titles, labels and explanatory copy if a requested design leaves wording open. Preserve user text exactly; put each visible string in its text field. Quote intended display strings inside captions using JSON escapes for BOTH opening and closing quotes. Do not add unrelated lettering. Stay within the response budget by reducing item count and prose; always close every string, object and array. No comments or Markdown fences.
Complete valid example for {{width}}x{{height}}:
{"summary":{"meta":["poster"],"canvas":"single","caption":"A cream travel poster with a blue mountain and a title."},"items":[{"label":"text","caption":"Large navy uppercase title.","text":"EXPLORE","bbox":[-{{half_height}},-{{half_width}},0,{{half_width}}]},{"label":"object","caption":"A flat blue mountain illustration.","bbox":[0,-{{half_width}},{{half_height}},{{half_width}}]}],"bbox_coord_space":"abs_centralized","bbox_format":"yxyx"}N)ÚEDITINGÚCLASSICÚ
STRUCTURED© ó    úAC:\Users\johnn\OneDrive\Documents\AI\models\moss_image\prompts.pyú<module>r      s!   ðð W€ð ~€ð~€
€
€
r   