Nano Banana 2 終極圖像生成指南!複製貼上專業技巧 | 內含 AI 提示詞與教學

Nano Banana 2 (Gemini 3.1 Flash Image) 現已在 Atlas Cloud 上線!圖像生成提示詞與模板已可在 Playground 或透過 API 使用,每張圖像價格低至 USD0.056!

大家好!Nano Banana 2 現已重磅登場,Atlas Cloud 已以閃電般的速度完成部署。不僅影像品質與生成速度大幅躍升,還有無數令人驚艷的全新玩法等您探索。

還在為撰寫 AI 生成提示詞而煩惱嗎?今天我們整理了一份 Nano Banana 2 海內外玩法全攻略。從圖生圖到文生圖,從電商海報到社群媒體推廣,這份逐步教學都附上了可直接複製的提示詞。建議先收藏,再開始您的創作之旅!

I. 圖生圖:參考圖的全新玩法

風格遷移 / 風格模仿

  1. 真人照片轉手繪風格
漫畫風格Redraw this image, manga illustration, smooth and clean outlines, high contrast colors, neon gradient lighting, exquisite transparent eyes, layered hair, dynamic posture, delicate clothing textures, soft diffuse lighting, 8K, ultra-high detail, original ratio
暖色色鉛筆Recreate the image. colored pencil drawing. hand drawn, sketch, (warm atmosphere:1.1), soft lighting, healing style, rough paper texture, vibrant but pastel colors, cozy, illustration, intricate details, masterpiece, best quality
水彩畫Recreate the image. watercolor painting, wet-on-wet, ink wash, colorful splatter, dripping paint, soft edges, ethereal, artistic style, liquid texture, flowing colors, white background, masterpiece
塗鴉/線稿Recreate the image. simple line art, realistic pencil sketch, doodle, stick figure style, flat lines, clean background, black and white, vector art, cute, childish drawing, abstract, few details, thick lines
素描Recreate the image. charcoal drawing, grayscale, monochrome, graphite, cross-hatching, detailed texture, light and shadow, rough sketch, academic art, high contrast

Photo to Hand-drawn.jpg

  1. 照片編輯

夕陽氛圍:Recreate the image. (golden hour:1.4), (sunset atmosphere:1.3), dramatic lighting, cinematic lighting, warm orange light, rim light, silhouette, lens flare, volumetric lighting, tyndall effect, emotional, dreamy, masterpiece, best quality

Sunset_20260228113810_436_61.jpg

夜晚模式:將其變為夜晚場景

Night_20260228114539_438_61.jpg

  1. 人像攝影

創意時尚雜誌風格:High-resolution studio portrait, close-up, retaining the reference image's model clothing, facial features, hairstyle, and glasses. White background with no identifiable elements. Her face is slightly turned, but her gaze is direct, intense, and focused. Her entire face is composed of puzzle pieces—each clearly visible with clean edges and a delicate 3D feel. She gently holds a piece she just removed from her cheek with her fingers, revealing a hollow black void underneath. Lighting is focused and directional, creating almost no shadows but accurately outlining the skin and the 3D puzzle structure. Overall atmosphere is surreal, precise, and highly realistic. Fashion magazine-style title "Atlas Cloud" at the top (behind the subject).

Magazine_20260228115210_440_61.jpg

冬日氛圍:Under a single lamppost, the man attached in the uploaded reference image ( Keep the face of the person 100% accurate from the reference image ) becomes a figure of solitude. The charcoal Dior Homme coat catches the warm glow of light as snow spirals through the dark. His scarf is wrapped high, Gucci gloves clutch the umbrella, and his gaze drifts upward toward the falling flakes. The surrounding night is blurred and dreamlike, a cinematic scene suspended between quiet introspection and timeless style.

image (84).png

專業設計 / 教育 / 圖表

3D 玻璃質感 Logo:3D embossed glossy contour render of center-aligned [BRAND, the reference pic] on a flat surface, perfectly centered composition with ample negative space surrounding the object for a premium minimalist aesthetic. Monochromatic [Purple] palette with soft tonal gradients. The object is defined by a raised, smooth, liquid-like glass bezel or chrome rim, creating a blind emboss effect where the interior matches the background. Matte surface finish with fine film grain or noise texture overlay. Soft diffuse lighting, strong specular highlights on the rounded edges, top-down view.

logo3d.png

電商應用

純白背景的多角度產品展示

Based on the uploaded reference image, extract the glasses from the reference pic and create a split composition showing four different angles (front, side, three-quarter, top-down) of the glasses, all against a clean white background. Product catalog style, uniform lighting, high resolution, 8k.

image.png

II. 文生圖:從零開始創作

海報設計

油畫風格水果海報:Vertical poster. Hand-drawn style, flat illustration, all text is left-aligned. Tiny text at the very top. "2026" on the left and right, "NAICHA DESIGN" in the middle. Theme "Plant Discovery Diary" in small English serif font on the top left (three lines). Tiny Chinese Songti text left-aligned at the bottom, followed by Chinese subtitle "葡串紫晶" in small Songti font. Small English text "Grape, Purple Crystal Cluster" and "Purple clusters condense sweetness, sweet juice moistens the throat". Smallest English text at the very bottom: "Not just a bunch of orbs, but a cluster of gems. Bathed in violet hues, it holds rich sweetness - plump flesh bursting with nectar, mellow and fragrant, carrying the warmth of autumn in every sip, gently. That fruits embody." Left-aligned layout. Overall style: Flat illustration, white background for text section. The image features Impressionist-style brushstrokes, Post-Impressionism, close-up, abstract and minimalist lines, soft texture, soft focus effect, creating a dreamy and ethereal atmosphere. Light spot brushes, bright and vivid colors, lavender mist blooming, soft light and shadow.

Gemini_Generated_Image_acau3kacau3kacau.png

真實感攝影

垂直螢幕,清新真實的照片。場景是一個略顯雜亂的房間。全身鏡旁站著一個像洋娃娃般的女孩,留著及腰長髮,戴著髮夾,穿著義大利 T 恤和格子披肩。衣服輕薄且有層次感。修長的指甲在閃光燈下閃閃發光。她正拿著「Line Puppy」手機殼進行鏡像自拍。快門速度產生的動態模糊。半身視角展示女孩的整體姿勢。溫暖的黃色調營造出溫馨的氛圍。

Gemini_Generated_Image_uhlnrzuhlnrzuhln.png

超寫實電影感街頭人像,一名年輕女子穿著亮藍色羽絨服,靜止在繁忙的城市人群中,與相機進行強烈的眼神交流,淺景深,動態模糊的行人在她身邊穿梭,背景是帶有交通燈和高樓大廈的都市,冷色調調色,自然的皮膚質感,柔和的陰天光線,85mm 鏡頭,f/1.8,細節豐富,主體對焦清晰,動態模糊,編輯攝影,8k。

Gemini_Generated_Image_m6a10rm6a10rm6a1.png

紀錄片風格場景,男性,皮膚白皙的青年,表情空靈憂鬱。黑色長髮在風中飄動,髮梢掛著玻璃珠。身穿華麗厚重的藏族服飾,佩戴銀飾。右手抬起,掌心面向鏡頭,眼神深邃。背景是藍天與雪山,經幡在空中飄揚。淺景深,膠片顆粒質感,動態美學,朦朧美學,自由畫風,先鋒藝術,細節豐富,高端質感,傑作。柔和朦朧的氛圍,精緻的光影,復古 CCD 濾鏡,輕微晃動模糊,暗部模糊。比例 9:16。

Gemini_Generated_Image_scfhw4scfhw4scfh.png

超現實電影/遊戲場景

地圖:

A vintage fantasy map showing the silhouette of a [subject] embedded in terrain features, rendered in hand-drawn ink and parchment-style textures. Mountains, forests, and rivers align to form the shape, colored in faded [color1] and [color2].

Gemini_Generated_Image_1lorf01lorf01lor.png

III. 進階:如何撰寫讓 Nano Banana 2 精準執行的提示詞?

看完上述範例,想創作屬於自己的傑作嗎?您只需要掌握這個 通用提示詞結構

[鏡頭/拍攝方式] + [主體] + [環境] + [光影] + [構圖] + [風格] + [畫質修飾詞]

  • 鏡頭/拍攝方式

    • 基礎選項:特寫 (Close-up)、近景 (Near shot)、中景 (Medium shot)、全身照 (Full shot)、遠景 (Long shot)、俯視 (Bird's eye view)、頂視 (Top-down view)、低角度 (Low-angle view)。
    • 創意選項:無人機視角 (Drone view)、第一人稱視角 (POV)、微距鏡頭 (Macro lens)、魚眼鏡頭 (Fisheye lens)、移軸效果 (Tilt-shift effect,微縮模型感)、望遠鏡視角 (Telescope view)、監控視角 (CCTV view)。
  • 主體: 穿著 JK 制服的女孩、一瓶深藍色香水、一座廢棄的未來城市。

  • 環境: 陽光下的櫻花樹下、雨夜街道、長滿青苔的森林。

  • 光影: 柔和自然光、強烈的霓虹燈、戲劇性的側光、丁達爾效應 (Tyndall effect)。

  • 構圖:

    • 基礎:三分法 (Rule of thirds)、居中構圖、對稱構圖、對角線構圖。
    • 進階:引導線 (利用道路/河流)、框架構圖 (透過門/窗)、留白 (Negative space)、景深 (前景模糊,背景清晰)、螺旋構圖。
  • 風格:

    • 藝術:扁平插畫、印象派油畫、水彩、素描、浮世繪、水墨、厚塗。
    • 數位/3D:3D 渲染、C4D 渲染、Octane 渲染、黏土動畫風格、像素藝術、體素藝術 (Voxel art)。
    • 攝影/電影:黑白攝影、電影感、寶麗來、紀錄片攝影、超現實主義。
    • 文化/類型:吉卜力風格、賽博龐克 (Cyberpunk)、蒸汽波 (Vaporwave)、北歐神話、洛可可藝術。
  • 畫質修飾詞: Masterpiece (傑作), high quality (高品質), incredible details (令人難以置信的細節), 8k, photorealistic (照片級真實)。

進階小技巧

  1. 提示詞越詳細,生成結果越精準。
  2. 嘗試組合不同的元素。
  3. 建立自己的提示詞庫,方便隨時調用。

IV. Atlas Cloud:駕馭 Nano Banana 2 的最佳平台

Atlas Cloud 核心優勢

高性價比與極致速度

  • 具競爭力的定價策略:透過優化價值,提供最具成本效益的生成體驗。
  • 同時體驗主流模型:無需訂閱多個平台。

工作流程整合與 API 生態系 專為無縫融入技術開發流程並支援下游任務而設計。

  • 靈活的協作工作流:支援在同一環境中與其他生成模型平行協作。
  • 高可用性交付成果:生成結果易於進行後製或二次編輯。
  • 全功能 API 存取:提供標準開發者介面,實現商業流程自動化與深度應用整合。

如何在 Atlas Cloud 上使用?

方法一:直接於平台使用

Nano Banana 2

方法二:透過 API 使用
第一步:取得您的 API Key

控制台 中建立並複製您的 API Key。

Guidance1.jpgGuidance2.jpg

第二步:參考 API 文件

請參閱我們的 API 文件,了解端點、請求參數與身份驗證方式。

第三步:執行您的第一次請求 (Python 範例)

Nano Banana 2 (文生圖) 為例:

plaintext
1import requests
2import time
3
4# 第一步:開始影像生成
5generate_url = "https://api.atlascloud.ai/api/v1/model/generateImage"
6headers = {
7    "Content-Type": "application/json",
8    "Authorization": "Bearer $ATLASCLOUD_API_KEY"
9}
10data = {
11    "model": "google/nano-banana-2/text-to-image",
12    "aspect_ratio": "16:9",
13    "enable_base64_output": False,
14    "enable_sync_mode": False,
15    "output_format": "png",
16    "prompt": "cyberpunk megacity at night, heavy rain, neon lights reflecting on wet streets, dense futuristic skyscrapers, giant holographic advertisements, flying vehicles in the sky, thick fog and atmospheric haze, moody cinematic lighting, dark dystopian atmosphere, ultra detailed, cinematic composition, 35mm film still, volumetric lighting, cyberpunk aesthetic",
17    "resolution": "2k"
18}
19
20generate_response = requests.post(generate_url, headers=headers, json=data)
21generate_result = generate_response.json()
22prediction_id = generate_result["data"]["id"]
23
24# 第二步:輪詢取得結果
25poll_url = f"https://api.atlascloud.ai/api/v1/model/prediction/{prediction_id}"
26
27def check_status():
28    while True:
29        response = requests.get(poll_url, headers={"Authorization": "Bearer $ATLASCLOUD_API_KEY"})
30        result = response.json()
31
32        if result["data"]["status"] == "completed":
33            print("Generated image:", result["data"]["outputs"][0])
34            return result["data"]["outputs"][0]
35        elif result["data"]["status"] == "failed":
36            raise Exception(result["data"]["error"] or "Generation failed")
37        else:
38            # 處理中,等待 2 秒
39            time.sleep(2)
40
41image_url = check_status()

最新模型

一個 API,暢享全模態 AI。

探索全部模型

Join our Discord community

Join the Discord community for the latest model updates, prompts, and support.