Style Transfer Comparison Prompt 画像プロンプト

コピー用画像プロンプト

画像モデルの追従性を高めるため、英語のプロンプト本文を使うのがおすすめです。

Use [STYLEREFERENCEIMAGE] as the visual style source.
Generate [NEWSUBJECTOR_SCENE] in the same visual language.
Preserve: color palette, brush/texture behavior, lighting mood, shape rhythm, level of abstraction, edge treatment.
Do not preserve: exact characters, copyrighted symbols, private logos, or the original composition.
Output layout: one main image plus three small variation thumbnails showing different composition options.
Scene requirements: [SUBJECTDETAILS], [BACKGROUND], [MOOD], [CAMERAORARTSTYLE].
Quality controls: coherent style match, clear subject silhouette, no muddy blending, no accidental text, no overfitted copy of the reference image.
Aspect ratio: 16:9 or 4:5 depending on platform.

使う場面

• Use this for social-ready AI image creation, visual exploration, campaign moodboards, or creative production drafts.
• Best for Nano Banana Pro, GPT Image, Midjourney, Recraft, Grok Imagine, or similar image models depending on the listed platforms.
• Replace the bracketed variables before generation.

置き換える入力

• [BRANDNAME], [PRODUCTNAME], [TITLE], [STYLEREFERENCEDESCRIPTION], or other bracketed variables.
• Optional input image or style reference where the prompt asks for one.
• Target aspect ratio and publishing channel.

品質ガードレール

• Keep identity, layout, and text constraints explicit.
• Do not copy a living artist, brand, poster, or private person's likeness without rights.
• Run multiple generations and select for prompt adherence, not just aesthetics.
• For commercial use, verify logo, face, wardrobe, and typography details manually.

Xでのトレンド信号: https://x.com/ComfyUI/status/2034753625340387395
This AIDreamHub version is rewritten from the observed trend direction and is not a verbatim copy of the X post.