Dual-panel AI portrait comparison

Objective: To create a side-by-side comparison image showcasing two AI-generated close-up portraits of the same young East Asian woman, highlighting subtle facial differences between the two versions. Canvas: Wide horizontal 2-panel layout, approximately 1200×525 pixels, with a thin vertical dividing line in the center. No border, no title, no watermark. Layout: Uses two completely equal panels. Left panel: The first portrait, labeled in small white text near the bottom right corner.{argument name="left label" default="1st time"} Right panel: The second portrait, marked in small white text near the bottom right corner.{argument name="right label" default="second time"} Labels should use a clean sans-serif font, white text with a subtle dark shading to ensure readability. Main details: Each panel displays a compactly cropped frontal portrait of the head and face, with the main subject being {argument name="subject" default="A young East Asian woman"} She had long, fluffy hair.{argument name="hair color" default="Black hair"} The portraits should complement the face shape, featuring a smooth forehead, naturally dark eyebrows, almond-shaped dark eyes looking directly at the camera, a small, straight nose, softly closed lips, and a calm, natural expression. The face should occupy most of the space on each panel, cropped close to the top of the forehead and around the chin, with hair extending to the sides. The two portraits should look very similar but not identical: eye shape, cheek fullness, nose bridge, mouth shape, and facial proportions should differ slightly. Visual style: Realistic AI portrait photography, using soft, cinematic lighting, shallow depth of field, smooth skin texture, soft warm skin tones, a dark blurred background, and a slightly soft-focus effect. The overall atmosphere should be quiet and serious. The left portrait should remain slightly soft and understated, while the right portrait should be slightly more refined, warmer, and more symmetrical. Constraints: Include 2 portraits and 2 text labels. Do not add additional descriptive text, UI elements, logos, or decorative graphics. Maintain direct facial contrast in the composition, with both faces centered and proportionally similar.