Four-Model Anime Poster Comparison

Goal: Create a wide social-media comparison graphic showing four AI-generated fictional anime movie posters side by side, evaluating Japanese title and release-date rendering accuracy.

Canvas: 1200 × 625 px horizontal image with a dark charcoal background. Add a top header area about 45 px tall, then four equal vertical poster panels beneath it separated by narrow dark gutters.

Header text: At the top left, set the large Japanese headline {argument name="headline text" default="画像モデル対決：同じキービジュアル・プロンプトを4モデルで生成（架空映画ポスター）"}. Below it in smaller gray Japanese text, add {argument name="subtitle text" default="評価軸＝日本語タイトル/公開日の描画精度。全モデル同一プロンプト・9:16"}.

Layout: Use exactly 4 poster columns, each with a thin dark label bar at the top. The four labels, from left to right, must read: 1) “gpt-image-2｜文字の完璧”, 2) “Nano Banana Pro｜字◎/日付2重”, 3) “Nano Banana 2｜字○/題2重”, 4) “Seedream 5.0｜誤字&改行”. Each poster is a vertical 9:16 anime film key visual.

Poster subject and style: All four posters depict the same fictional Japanese anime romance/drama film set at sunset by the sea. Use a luminous, emotional anime movie-poster style, soft painterly lighting, warm orange and pink sunset clouds, sparkling firefly-like light particles, coastal hills, a distant seaside town, ocean horizon, birds in the sky, and a melancholic young woman with windblown dark brown hair and large reflective eyes. Keep each poster visually similar but slightly different in crop and composition to suggest outputs from different image models.

Panel details: Panel 1 shows the girl in upper-left close-up turning toward the viewer, with the sea and town below; it has disciplined typography with the title once near the middle, English subtitle below, and release date once at the bottom. Panel 2 shows a closer portrait of the girl over a hillside town and sea; intentionally duplicate the release-date line at the bottom twice while keeping the title once. Panel 3 shows the girl centered above a blurry seaside town; intentionally duplicate the Japanese title twice, one near the middle and one lower on the poster, while keeping one release-date line at the bottom. Panel 4 shows a close-up face on the right with ocean and town below; make the Japanese title appear awkwardly broken across lines and slightly wrong-looking, with the release date line centered near the lower middle.

Poster text content: The fictional film title is {argument name="movie title" default="サヨナラの向こうがわ"}. The English subtitle is {argument name="english subtitle" default="BEYOND GOODBYE"}. The release text is {argument name="release date text" default="2027年7月16日 全国ロードショー"}. Use elegant Japanese Mincho-style title lettering and small spaced serif caps for the English subtitle. Make the first panel the cleanest and most typographically correct.

Constraints: Maintain exactly four panels and exactly the specified label bars. Do not add logos, watermarks, extra panels, extra model names, or unrelated text. The final image should look like a tweet-attached benchmark image comparing layout discipline across four image models.