# GPT Images用プロンプト セミナーで見せる画像生成プロンプト集。 目的は「きれいな画像」を作ることではなく、**LPの申し込み理由が伝わる画像**を作ること。 ## 使い方 1. 参照画像がある場合は、人物の顔・雰囲気の参照だけに使う。 2. 画像内に文字、ロゴ、看板、YouTube UIを入れない。 3. LP側で見出しを置くため、左右どちらかに余白を作る。 4. 生成後は、手・顔・文字化け・不自然なツヤを必ず確認する。 --- ## 共通の禁止指示 どのプロンプトにも最後に足す。 ```text no text, no logo, no readable sign, no YouTube UI, no microphone, no official trademark, realistic hands, natural face, no AI gloss, editorial photo, negative space for web copy ``` --- ## 1. ナオミ焼肉 / ヒーロー 狙い: 「高級焼肉」ではなく、**店主に会いに行く焼肉店**に見える画像。 ```text Use the reference image only for Naomi's identity and face continuity. Create a cinematic hero image for a fictional Japanese yakiniku landing page. The concept is: people come to this restaurant to meet Naomi, the owner, not only to eat expensive beef. Naomi stands behind a dark charcoal grill counter, close enough to feel like she is welcoming one guest. Warm flame and soft smoke in the foreground. Black premium restaurant interior, small counter seats, warm wood, ceramic plates, lived-in but refined mood. Naomi is not posing like an ad model; she looks like the owner about to serve the first slice. Asymmetrical composition, Naomi on the right side, dark negative space on the left for web headline. Documentary editorial photography, realistic light, slight film grain. no text, no logo, no readable sign, no YouTube UI, no microphone, realistic hands, natural face, no AI gloss ``` ### 講義で見せる3枚構成 ```text 1. 店主: Naomi at a small yakiniku counter, looking at the guest, one hand near tongs, warm smoke, approachable but premium, no text. 2. 店内: small counter yakiniku restaurant, 8 to 10 seats, black stone, warm wood, soft grill smoke, lived-in not generic luxury, no people facing camera, no text. 3. 一枚目: close documentary shot of the owner placing the first slice of marbled beef on a grill, tongs, flame, smoke, ceramic plates, imperfect real texture, no text. ``` --- ## 2. 俳優ナオミ / プロフィール 狙い: 「美しいポートレート」ではなく、**仕事を頼めそうな俳優の空気**を作る。 ```text Use the reference image only for Naomi's identity and face continuity. Create a refined portrait for Naomi's fictional actor profile landing page. Naomi stands in a quiet dark studio with dramatic side light. The mood is not idol-like and not cute. It should feel like an actor's official profile: calm expression, controlled posture, strong eyes, subtle tension before a performance. Black and ivory palette, minimal background, slight curtain or shadow movement, large negative space for a headline and profile text. Editorial magazine photography, mature and restrained, realistic skin texture, natural hands, no over-smoothed AI look. no text, no logo, no readable sign, no YouTube UI, no microphone, realistic hands, natural face, no AI gloss ``` ### 差し替え用カット ```text Naomi sitting near a rehearsal room window, script pages on a table, soft side light, quiet concentration, actor profile mood, documentary editorial photo, no text, no logo, realistic hands, natural face ``` --- ## 3. ナオミ隊長 / 秘境ナイトツアー 狙い: 「森で遊ぶ」ではなく、**ナオミ隊長についていきたくなる体験**に見せる。 ```text Use the reference image only for Naomi's identity and face continuity. Create a cinematic hero image for a fictional night adventure tour landing page. Naomi is the expedition leader guiding guests into a hidden place at twilight. She stands at the edge of a forest path with a warm lantern and a small map, looking back as if saying "follow me". She wears a practical khaki guide shirt, olive or tan utility vest, canvas belt or shoulder strap, small scarf, and simple outdoor hat. The mood is adventurous but not cute. Show concrete tour details: lantern light, narrow wooden bridge in the distance, dark green leaves, mist, a safe walking path, faint signs of a hidden valley beyond the trees. The visitor should feel: if Naomi leads, I can go there. Asymmetrical composition, Naomi on the right, large dark negative space on the left for the web headline. Premium dark website mood, realistic face and hands, natural outdoor light. no text, no logo, no readable sign, no official trademark, no YouTube UI, no microphone, no AI gloss ``` ### 講義で見せる3枚構成 ```text 1. 出発: Naomi holding a lantern at the meeting point, guests are implied by the camera angle, twilight forest, safe but exciting, no text. 2. 途中: close shot of boots and lantern crossing a narrow wooden bridge, leaves, mist, handrail, realistic night tour detail, no text. 3. 到着: hidden valley or small waterfall softly lit by lanterns, Naomi slightly ahead pointing the way, guests feel guided, no text. ``` --- ## 失敗したときの直し方 ### 文字が出る ```text Remove all text, signs, letters, logos, labels, watermark-like marks. The image must contain no readable characters. Put all copy in HTML later. ``` ### 顔がAIっぽい ```text Make the face more natural and less glossy. Preserve identity from the reference image, but use realistic skin texture, asymmetry, natural expression, documentary lighting, no doll-like face. ``` ### 手が崩れる ```text Hands must be realistic with correct fingers. If hands are difficult, keep hands partially hidden by tools, lantern, tongs, table edge, or natural motion blur. ``` ### きれいすぎて広告っぽい ```text Make it more lived-in and documentary. Add slight film grain, imperfect real texture, natural shadows, practical objects, and a real sense of place. Avoid generic luxury advertising. ``` ### LPに文字を置く場所がない ```text Keep the subject on the right third of the image and leave clean negative space on the left for a large Japanese headline. Do not place important faces or hands in the left text area. ```