Explore. Create. Go viral.

Create a hyper-realistic split-screen comparison image showing the same young woman underwater in two different AI-rendered interpretations. The composition is vertical with two equal rounded-rectangle panels side by side, separated by a narrow dark divider. Both panels show the subject swimming underwater in clear tropical blue water, smiling toward the camera with natural body motion and sunlight filtering from the surface above.

In the left panel, the woman is closer to the camera and wears a bright pink string bikini. She has long dark hair gathered in a high ponytail, large round silver eyeglasses, and medium hoop earrings. Her face is very close and crisp, and her body floats vertically in the water with a friendly, open smile. The water is bright turquoise and simple, with a clean pool-like or open-water clarity.

In the right panel, the same woman is shown farther back in a more full-body underwater swimming pose. She wears a black one-piece swimsuit and her hair flows freely in the water. The environment includes a sandy seabed, a few small reef fish, and visible coral formations near the lower right. The look is more environmental and scenic while still preserving the same facial identity and cheerful expression.

Both panels should include realistic underwater light rays, suspended bubbles, surface shimmer, and gentle distortion from being submerged. Near the lower area between the panels, include small prompt text in Spanish reading: "Prompt: La chica está buceando en el mar". At the bottom of the left panel place the label "SEEDREAM 4" in a warm yellow-orange tone. At the bottom of the right panel place the label "NANO-BANANA" in the same color style. The overall result should feel like a clean social media comparison card testing how two image models handle underwater human portraits, identity consistency, accessories, and environmental realism.

Prioritize realism, underwater clarity, comparison readability, consistent identity across both sides, and believable swimming posture in each model panel.