首頁
圖片生成
影片生成
靈感社區
生成記錄
S

Saman | AI

@Samann_ai

GPT Image

超寫實街頭風格人像攝影:真人擁抱卡通版自己

提示詞

{ "title": "Hyper-real street portrait: real person hugging a clearly CARTOONY version of themselves", "inputs": { "user_photo": "REPLACE_WITH_USER_FACE_PHOTO_URL_OR_IMAGE_ID", "style_reference_image": "REPLACE_WITH_CARTOON_STYLE_REFERENCE_URL_OR_IMAGE_ID" }, "prompt": "Create a cool, hyper-real, street-style portrait in a real urban street environment at golden hour (natural light, city bokeh, shallow depth of field). Use the provided USER PHOTO to preserve the person’s identity ONLY for the photoreal character.\n\nKEY GOAL: The difference between REAL and CARTOON must be very obvious.\n\nScene & composition:\n- Vertical portrait, waist-up preferred, centered.\n- A photoreal, hyper-real version of the person stands behind.\n- In front: the person’s CARTOON VERSION in a premium 3D animated style strongly matching the STYLE REFERENCE IMAGE.\n- The photoreal person hugs the cartoon from behind: one arm gently around the cartoon’s neck/upper chest in a friendly embrace (not choking), the other hand lightly resting on the cartoon’s shoulder.\n- Both smiling warmly and looking directly into the camera.\n\nPhotoreal person requirements:\n- 100% realistic human: visible skin pores, natural imperfections, realistic hair strands, authentic fabric weave.\n- Realistic streetwear outfit that complements the cartoon’s colors.\n\nCartoon version requirements (MAKE IT MORE CARTOONY):\n- Must look like a stylized animated character, NOT realistic.\n- Exaggerated proportions: slightly larger head, bigger expressive eyes, simplified nose/mouth, softer facial planes.\n- Smooth skin with NO pores, NO real-world micro-textures.\n- Stylized hair: clean clumped curls/strands with animated volume (not individual photoreal strands).\n- Use toon-like material response: soft highlights, clean shading, subtle subsurface glow, crisp edges.\n- Keep clear resemblance to the user (overall face shape cues), but do NOT make it realistic.\n- Outfit & accessories should follow the STYLE REFERENCE IMAGE vibe closely (boho/patchwork streetwear, layered handcrafted details, playful colors) with clearly animated fabric detail.\n\nLighting & camera:\n- Photoreal character uses natural golden-hour lighting with realistic shadows.\n- Cartoon character uses matching scene lighting BUT rendered with stylized shading (animated look).\n- 50mm lens look, f/1.8, cinematic grade, sharp focus on both faces.\n\nImportant:\n- Both characters must be in the same frame and clearly different in rendering style.\n- The cartoon must NEVER look photoreal.", "negative_prompt": "cartoon looks realistic, pores on cartoon skin, photoreal cartoon hair, uncanny valley, same rendering style for both, low quality, blur, noise, extra fingers, deformed hands, duplicated limbs, aggressive choking, violence, bruises, blood, text, watermark, logo, caption, frame, heavy artifacts", "params": { "aspect_ratio": "4:5", "quality": "ultra", "stylization_strength": 0.85, "identity_preservation_strength": 0.9, "reference_strength": 0.9, "depth_of_field": "strong", "seed": 12345 }, "tuning_tips": [ "If your cartoon still looks too real: increase stylization_strength to 0.95 and add more 'exaggerated proportions, big eyes, smooth skin, toon shading' phrases.", "If the cartoon stops resembling you: keep the REAL person locked to your user_photo (identity_preservation_strength high), but for the cartoon rely more on the style_reference_image and reduce any face-detail realism cues (remove words like 'highly detailed skin' from cartoon section).", "If your tool supports separate controls per subject: set HIGH identity for photoreal, MEDIUM identity for cartoon, HIGH style for cartoon." ] }

更多靈感

瓷白肌理東亞少女唯美肖像寫真VELUNE榛果可可醬:現代玻璃高樓城市中心的巨型戶外看板視覺超現實混合媒介拼貼:黑白影像中一抹亮黃的單車少女童趣手繪裝飾民俗扁平插畫:可愛塗鴉元素作品色彩炸裂的墨西哥塔可街頭風食譜卡設計
光影流轉的奢華腕錶廣告海報——TIME MOVES LIGHTON REPEAT:將回憶、城市與夢想寫進歌單的直式生活海報貝克漢黑白海報:高對比肖像,展現真實力量水墨黑白人像:高反差墨跡與戲劇性光影皮克斯風格香蕉男角色:自信眼神與閃亮微笑的3D插畫
用作提示詞用作參考圖
圖片影片靈感記錄