同一角色同 prompt 三图模型横评:Nano Banana Pro 对阵 GPT Image 2.6
gizakdag · x · 2026-11-12
In Figma Weave, the author connected one character reference image to three image models at once — Nano Banana Pro, GPT Image 2.6 and Nano Banana 2.1 — with the same reference and the same prompt ("turn him a bear"), then compared results side by side.
All three preserved the red-and-blue line style but differed in body shape and details. The author says this parallel comparison made choosing a model easy — a workflow trick worth borrowing for any AI image pipeline.
「多模态」频道最新
- 用 Claude Fable 5.2 把《霹雳游侠》片头重制成 Netflix 风格剧集开场 — justin_hart · 2026-11-12
- Gemini 4 Carbon 花 6 小时从零搭出含鱼虾蜗牛的 3D 水族箱 — sunjiao123sun_ · 2026-11-12
- 小团队开源 Awaken:LoRA 微调把半双工语音 LLM 改造成全双工对话 — alexcovo_eth · 2026-11-12
- MiniMax H3 Image 主打电商图：草图 10 秒生成 4K 全场景素材 — MiniMax_AI · 2026-11-12
- Kaigen 3D 工具演示:一个人就能做出完整 3D 场景 — gdechichi · 2026-11-12
- 一人用 Tripo+Fable+自研引擎做出丧尸群像关卡 — gdechichi · 2026-11-12