§ ’Þj5 ãó2—dZdezZdezZdezZedzZdS)a©Preserve the user's intent, language, exact supplied display strings, and explicit exclusions. When a design needs writing and the wording is open, author complete suitable titles, labels and body copy. Quote every intended visible string exactly; never leave placeholder copy or unspecified text areas. Do not add lettering to a purely visual request. Return only the enhanced prompt, with no explanation or Markdown fences.z‘Write a detailed image-generation caption for Fibo 2.0. Describe the finished scene, subjects, composition, materials, light, palette and style. zâWrite a concise Fibo 2.0 editing instruction grounded in the supplied references. State changes and preservation constraints. Number sources in supplied order as , , etc. Do not invent additional references. a$Write a complete valid Fibo 2.0 scene JSON object. Supported scene fields are short_description (string), objects (list of descriptive objects), background_setting (string), lighting (object), aesthetics (object), photographic_characteristics (object), style_medium (string), context (string), artistic_style (string), and text_render (list of objects with a text string). Omit unused sections. Use semantic spatial descriptions in objects; there is no numeric-coordinate or bounding-box schema. Assign each subject and string to one owning element without duplicate objects. Every intended visible string must appear in text_render with exact text, and can include a location description. In descriptions quote any mentioned display strings with JSON-escaped double quotes. Close every quote, object and array; shorten descriptive detail to fit the response budget instead of truncating. Complete example: {"short_description":"A cream travel poster showing blue mountains and the title \"EXPLORE\".","objects":[{"description":"A layered blue mountain silhouette in the center"}],"background_setting":"Warm cream paper","style_medium":"Editorial poster","text_render":[{"text":"EXPLORE","location":"Large title above the mountains"},{"text":"Find your next trail","location":"Small subtitle below the mountains"}]} aô For this editing task, put the scene object described above inside structured_caption. The final response MUST have exactly these two required top-level fields: structured_caption and edit_instruction. edit_instruction is a nonempty string stating the requested change and what to preserve. Reference images in supplied order as , , etc. Complete final-response example: {"structured_caption":{"short_description":"The source portrait with a red leather jacket"},"edit_instruction":"Change the jacket in to red leather. Preserve the person's face, pose, lighting and background."} Before returning, verify that BOTH top-level fields are present and that every intended visible string appears in structured_caption.text_render. N)Ú_COMMONÚCLASSICÚEDITÚ STRUCTUREDÚSTRUCTURED_EDIT©óúr s]ðð }€ð bðelñ l€ðpðszñz€ððñ € ð ð ñ€€€r