I need an individual "embroidered travel patch" premium edited poster for each travel photo I upload, without combining multiple images, and with each photo output separately. The overall composition should be a 3:4 vertical comparison of the original image. The upper part should clearly retain a substantial area of the original photo, while the lower part should feature a single embroidered travel subject extracted from the same photo. Both upper and lower parts typically occupy about 50% each, which can be slightly adjusted based on the photo's aspect ratio, but the comparison with the original image must not be removed. The upper part should faithfully retain the subject's identity, quantity, posture, structure, viewing angle, environmental relationship, authentic materials, natural light and shadow, and original colors in high fidelity. Only slight color adjustments typical of travel magazines and art publications should be applied. The upper part MUST remain an authentic photograph and must not be illustrated, have its scene replaced, subjects added or removed, or the structure of buildings and vehicles altered. If the original image ratio is not suitable for the horizontal area, it should retain its own ratio centered, and the pure color sidebars of the lower fabric background should be used to fill the gaps, without stretching, mirroring, or using blurred photos to fill the edges. The lower part should not embroider the entire photo completely. Only select the most recognizable core subject with the clearest outline from the photo, such as a tram, a boat, a bicycle, a hot air balloon, an animal, a person with a clear action, or a travel object. Retain the outer contour, orientation, posture, key structure, and iconic color blocks that define the subject's identity, while removing clusters of buildings, streets, sky, water surfaces, mountains, trees, and other complex backgrounds. The subject should use a scaled-down refinement or single-item display composition, occupying about 28%—52% of the lower area's width and 30%—50% of its height, usually placed in the middle or lower-middle. Approximately 45%—65% of continuous, complete, and unfragmented fabric open space should be retained around it. 2—5 short stitches can be added to suggest tracks, ripples, wind direction, ground, or direction of movement, but they must not develop into a complete background, nor should a second visual subject appear. The lower background should be a whole piece of real, visible twill, old canvas, coarse cotton, or linen. The fabric background color MUST be extracted from the original photo and adaptively change for each photo, such as gray-blue, terracotta pink, teal, smoky gray-blue, warm beige, or other low-saturation colors. Do not fixate on using the same off-white fabric for all photos. Use real satin stitch, split stitch, chain stitch, backstitch, and a small amount of loose thread to form the subject. The stitch direction MUST change with the subject's volume, structure, and movement: metal, wood, clothes, animal fur, wheels, boat hulls, balloons, or plants need to be expressed using embroidery threads of different directions, lengths, and pile heights. Local areas are allowed to have raised embroidery, slightly uneven stitches, misaligned edges, exposed backing, loose threads, thread ends, and creased fabric, making it appear like a genuinely hand-stitched embroidery on fabric, rather than a printed illustration or digital filter. Each piece should extract 4—8 embroidery thread colors from the original photo, retaining the most recognizable color relationships of the subject, then appropriately generalizing and softening them. Do not use fixed rainbow threads, neon gradients, or high-saturation clashing colors unrelated to the photo. The text is concentrated below the embroidered subject, using chain stitches with a lower pile height, small woven labels, or fabric embroidered text. By default, it includes a 1—3 English word title, a "PATCH NO. 01" field, and a brief observation sentence of 5—10 English words. The text content must be extracted from the real visible subject, action, or state in the photo and must be clearly readable...
