
Create a photorealistic 2x2 grid collage (four panels) in a modern indoor gym with mirrors, racks, dumbbells, and fluorescent ceiling lights. Each panel looks like a smartphone photo (slight wide-angle, realistic phone HDR, subtle compressi

{ "composition": { "layout": "Dual-panel vertical diptych", "symmetry": "Mirror-image structural composition with contrasting color palettes", "perspective": "High-angle, whimsical landscape with a central focal bottle", "central_focal_poin

Créez un design de texte en style dessin animé psychédélique et grotesque qui dit « GNARLY ». Arrangez les lettres en ligne horizontale droite. Chaque lettre doit être bosselée, fondue et suinter avec des couleurs vives et contrastées à pla

Create a visually stunning minimalist colored-pencil and wax-crayon artwork of [CITY], where the entire city is constructed from one single continuous hand-drawn line. The line begins as a tiny cultural symbol, travels through famous archit

A magical cinematic portrait of a young woman use image for reference with deep glowing under bioluminescent light, her face softly illuminated by shimmering stardust particles. Intricate braided hair woven into a regal crown, strands spark

{ "Subject": "A single fictional adult Japanese woman over 20 years old, sitting on a sofa bed in a bright room, wearing a black ribbed top and blue jeans, smiling while lightly touching her cheek with her hand. The core is the moment she

Misty light gauze tube top; loose clothing edges; chest lines softly revealed; half-hidden by layers of gauze curtains; walking slowly barefoot; skirt hem trailing on the ground; backlit rim lighting; a fleeting glimpse 雾感轻纱抹胸;衣缘松散;胸口线条柔和显

{ "task": "image_generation", "reference_image": { "role": "identity_reference_only", "instruction": "Use the uploaded reference image only to preserve the person's identity. Do not copy the reference image's original pose, clothing, light

Create a premium 4:5 vertical architectural travel artwork for [COUNTRY / CITY / LOCATION / LANDMARK]. TEXT-TO-IMAGE ONLY — generate the entire scene creatively from [COUNTRY / CITY / LOCATION / LANDMARK] alone. No reference image required