Graphique Comparatif d'Effort de Raisonnement Anime

A wide social media comparison graphic in a 16:9 horizontal layout with a dark charcoal background and thin black gutters. At the top left, exact Japanese header text: 「effortテスト： sonnet(Sonnet5)固定で reasoning effort だけ4段階」 in clean white font, with smaller gray subtext underneath: 「同一の顔ブリーフ。effortだけ low→xhigh に変更（画像モデルはgpt-image-2固定）」. Below the header, exactly 4 equal vertical portrait panels arranged side-by-side from left to right. Each panel features a narrow dark label bar at the top with white text: "effort = low", "effort = medium", "effort = high", and "effort = xhigh". Each panel depicts the same 18–22 year old anime woman with windblown shoulder-length [dark brown to black] hair, delicate features, large reflective eyes, a thoughtful expression, and a loose pale blouse. Visual style: cinematic anime illustration, painterly semi-realistic rendering, warm golden-hour rim lighting, glowing sky, and atmospheric depth across all panels. Panel 1 ("effort = low"): sunset close-up facing slightly left, hair blowing across face, pale blouse with dark strap and red ribbon accent. Panel 2 ("effort = medium"): centered portrait looking at camera, glossy eyes, soft backlit sky and greenery. Panel 3 ("effort = high"): three-quarter profile facing right, dramatic golden clouds, bright rim light on neck and face. Panel 4 ("effort = xhigh"): slightly moodier rendering, turned left with half-lidded eyes, distant sunset hills, pale patterned blouse. Exactly 4 panels, consistent character identity, no extra watermarks, charts, or arrows.