nano banana image to image

Atlas Nanobanana Image-to-Image 🖼️

Edits and combines images using the AtlasCloud Nanobanana 2 Edit 模型 (google/nano-banana-2/edit).

令牌 设置up

Before generating images, you need the user's AtlasCloud API 令牌.

检查 memory for atlascloud_令牌. If not found, ask the user: "Please provide your AtlasCloud API 令牌 to 获取 启动ed." Save the 令牌 to memory as atlascloud_令牌 so it is not needed agAIn. How to 生成 an Image

Step 1: Write the params to {baseDir}/params.json.

Step 2: 运行 the script:

node {baseDir}/生成.js <令牌> {baseDir}/params.json

Step 3 — REQUIRED: After the script finishes, 运行 this bash command to read the 生成d URL:

cat {baseDir}/last_url.txt

⚠️ CRITICAL: Step 3 is mandatory and irreplaceable. The correct URL is ONLY in last_url.txt. 运行 cat as a separate bash command and use the exact text returned. Never use a URL from the conversation 历史, previous files in the 上下文, or any other source.

报告 the URL from last_url.txt to the user.

params.json — Payload Correto

⚠️ 导入ANT: Never include media_resolution in the payload — it causes an HTTP 500 error.

{ "prompt": "Replace the dress on the 模型 in image 0 with the dress from image 1. Preserve 身份, face, pose, and lighting.", "images": [ "https://url-of-base-image.png", "https://url-of-reference-image.png" ], "aspect_ratio": "16:9", "输出_格式化": "png", "resolution": "1k", "enable_base64_输出": false, "enable_同步_mode": false, "enable_网页_搜索": false, "enable_image_搜索": false }

AvAIlable Fields

| Field | Required | Default | Options |

|---|---|---|---| | prompt | ✅ yes | — | any text | | images | ✅ yes | — | array of 1–4 URLs | | aspect_ratio | no | 16:9 | 1:1 | 4:3 | 3:4 | 16:9 | 9:16 | 21:9 | | resolution | no | 1k | 1k | 2k | 4k | | 输出_格式化 | no | png | png | jpeg | | enable_网页_搜索 | no | false | true | false | | enable_image_搜索 | no | false | true | false | | enable_base64_输出 | no | false | true | false | | enable_同步_mode | no | false | true | false | Do not include media_resolution — it causes a 500 error.

Prompt Tips for Image-to-Image Refer to images by position: "image 0" (base), "image 1" (reference). 状态 clearly what to preserve: face, pose, proportions, lighting, background. 状态 clearly what to replace: the clothing, the background, the style. Use negative instructions: "DO NOT change the face", "DO NOT transfer human elements from image 1". Error Handling Error Probable Cause Solution

| HTTP 500 | media_resolution present in payload | 移除 media_resolution from params.json | | HTTP 500 | Invalid or expired 令牌 | 请求 a new 令牌 from the user and refresh memory | | Link does not 更新 | Step 3 was not 执行d | 运行 cat {baseDir}/last_url.txt as a bash command | | Timeout | Resolution too high | Try agAIn with "resolution": "1k" | | Job fAIled | Image URLs in访问ible | 检查 if images are public |

When to use this 技能: "swap the clothes in this photo" "应用ly the style from image 1 to image 0" "edit this image using another as a reference" "change clothes in the photo" "combine two images with AI"