创意空间 - FreeGPTBanana

新图像

空白工作区

生成结果缩略图会显示在这里。

生成参数

提示词、参考图和生图参数。

Goal: Create a wide social-media comparison graphic showing an AI image generation experiment about reasoning effort. The graphic compares four anime-style portrait outputs side by side while keeping the character brief consistent. Canvas: Horizontal 1200×675 style image with a dark charcoal background, like a tweet attachment or experiment result slide. Use a clean, compact layout with thin black gutters between panels. Header text: At the top left, show the Japanese title 「effortテスト： sonnet(Sonnet5)固定で reasoning effort だけ4段階」 in white. Directly underneath, in smaller gray text, show 「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）」. Layout: Below the header, create exactly 4 equal vertical portrait panels arranged left to right. Each panel has a narrow dark label bar at the top with white text. The labels are exactly 4: “effort = low”, “effort = medium”, “effort = high”, and “effort = xhigh”. Subject details: In every panel, depict the same young anime woman, about 18–22 years old, with {argument name="hair color" default="dark brown to black"} windblown shoulder-length hair, soft delicate facial features, large reflective eyes, a slightly melancholic or thoughtful expression, natural skin tones, and a loose pale blouse. Keep her identity consistent across all 4 panels but allow slight variations in angle, expression, and lighting. Panel contents: Panel 1 “effort = low” shows her in a warm sunset close-up, facing slightly left of camera, hair blowing across her face, wearing a light blouse with a visible dark strap and subtle red ribbon accent. Panel 2 “effort = medium” shows her centered, looking toward the viewer with glossy eyes, soft backlit sky and greenery behind her, hair swept around her cheeks. Panel 3 “effort = high” shows her in a three-quarter profile, turned toward the right with blue eyes catching the light, dramatic golden clouds behind her, and stronger rim light on her face and neck. Panel 4 “effort = xhigh” shows her slightly more realistic and moody, turned left with half-lidded eyes, long dark hair blown across her face, warm sunset over distant hills, and a pale patterned blouse. Visual style: High-quality cinematic anime illustration, painterly semi-realistic rendering, warm golden-hour backlight, detailed hair strands, glowing sky, soft atmospheric depth, subtle filmic contrast, consistent color palette of amber sunlight, blue sky, and dark hair. The overall impression should be that all four outputs are similarly high quality with no obvious winner. Constraints: Use exactly 4 portrait panels and exactly the 4 labels listed above. Preserve the Japanese header text exactly. Do not add logos, watermarks, extra captions, charts, arrows, or additional panels.

参考图

这个 Prompt 需要你上传一张图，再开始生成。

模型

比例

分辨率

数量

质量

暂时无法获取价格