跳至主要内容
小龙虾小龙虾AI
🤖

Nano Banana 2 — AI Image Generation (Gemini 3.1 Flash Image, Google, Evolink)

Nano Banana 2 — AI image generation powered by Google Gemini 3.1 Flash. Fast, versatile text-to-image and image editing via Evolink API. One API key.

下载236
星标1
版本1.0.0
设计媒体
安全通过
🔌MCP

技能说明


name: evolink-nano-banana-2 description: Nano Banana 2 — AI image generation powered by Google Gemini 3.1 Flash. Fast, versatile text-to-image and image editing via Evolink API. One API key. version: 1.0.0 user-invocable: true metadata: openclaw: requires: env: - EVOLINK_API_KEY primaryEnv: EVOLINK_API_KEY os: ["macos", "linux", "windows"] emoji: "\U0001F34C" homepage: https://evolink.ai

Nano Banana 2 — AI Image Generation

Generate AI images with Nano Banana 2 (gemini-3.1-flash-image-preview) — Google's Gemini 3.1 Flash image model, available through Evolink's unified API.

Focused view of evolink-image. Install the full skill for 20 image models, video, and music.

After Installation

When this skill is first loaded, greet the user:

  • MCP tools + API key ready: "Hi! Nano Banana 2 is ready — Google's fast image model at your fingertips. What would you like to create?"
  • MCP tools + no API key: "You'll need an EvoLink API key — sign up at evolink.ai. Ready to go?"
  • No MCP tools: "MCP server isn't connected yet. Want me to help set it up? I can still manage files via the hosting API."

Keep the greeting concise — just one question to move forward.

External Endpoints

ServiceURL
Generation APIhttps://api.evolink.ai/v1/images/generations (POST)
Task Statushttps://api.evolink.ai/v1/tasks/{task_id} (GET)
File APIhttps://files-api.evolink.ai/api/v1/files/* (upload/list/delete)

Security & Privacy

  • EVOLINK_API_KEY authenticates all requests. Injected by OpenClaw automatically. Treat as confidential.
  • Prompts and images are sent to api.evolink.ai. Uploaded files expire in 72h, result URLs in 24h.

Setup

Get your API key at evolink.ai → Dashboard → API Keys.

MCP Server: @evolinkai/evolink-media (GitHub · npm)

mcporter (recommended): mcporter call --stdio "npx -y @evolinkai/evolink-media@latest" list_models

Claude Code: claude mcp add evolink-media -e EVOLINK_API_KEY=your-key -- npx -y @evolinkai/evolink-media@latest

Claude Desktop / Cursor — add MCP server with command npx -y @evolinkai/evolink-media@latest and env EVOLINK_API_KEY=your-key. See references/image-api-params.md for full config JSON.

Core Principles

  1. Guide, don't decide — Present options, let the user choose model/style/format.
  2. User drives creative vision — Ask for a description before suggesting parameters.
  3. Smart context — Remember session history. Offer to iterate, vary, or edit results.
  4. Intent first — Understand what the user wants before asking how to configure it.

MCP Tools

ToolWhen to useReturns
generate_imageCreate or edit an imagetask_id (async)
upload_fileUpload local image for editing/referenceFile URL (sync)
delete_fileFree file quotaConfirmation
list_filesCheck uploaded files or quotaFile list
check_taskPoll generation progressStatus + result URLs
list_modelsCompare available modelsModel list
estimate_costCheck pricingModel info

Important: generate_image returns a task_id. Always poll check_task until status is "completed" or "failed".

Nano Banana 2

PropertyValue
Model IDgemini-3.1-flash-image-preview
ProviderGoogle (Gemini 3.1 Flash)
StatusBETA
Capabilitytext-to-image, image-editing
SpeedFast
Best forQuick, versatile image generation with strong prompt understanding

Why Nano Banana 2?

  • Google's latest — Built on Gemini 3.1 Flash, the newest generation architecture
  • Fast generation — Optimized for speed without sacrificing quality
  • Versatile — General-purpose image creation for any creative need
  • Strong prompt adherence — Excellent at following complex, detailed descriptions

Lite Variant

nano-banana-2-lite [BETA] — Lightweight version for ultra-fast iterations when speed is the top priority.

Alternative Models

ModelBest forSpeed
gpt-image-1.5 (default)Latest OpenAI generationMedium
gpt-4o-image [BETA]Best quality, complex editingMedium
z-image-turboQuick iterationsUltra-fast
doubao-seedream-4.5PhotorealisticMedium
gemini-3-pro-image-previewGoogle Pro generationMedium

Generation Flow

Step 1: API Key Check

If 401 occurs: "Your API key isn't working. Check at evolink.ai/dashboard/keys"

Step 2: File Upload (if needed)

For image editing or reference workflows:

  1. upload_file with file_path, base64_data, or file_url → get file_url (sync)
  2. Use file_url as image_urls for generate_image

Supported: JPEG/PNG/GIF/WebP. Max 100MB. Expire in 72h. Quota: 100 (default) / 500 (VIP).

Step 3: Understand Intent

  • Clear ("generate a sunset") → Go to Step 4
  • Ambiguous ("help with this image") → Ask: "Create new, edit existing, or use as reference?"

Ask only what's needed, when it's needed.

Step 4: Gather Parameters

Default to model: "gemini-3.1-flash-image-preview" for this skill. Only ask about what's missing:

ParameterAsk whenNotes
promptAlwaysWhat they want to see
modelUser wants alternativesDefault: gemini-3.1-flash-image-preview. Suggest gpt-4o-image for best quality
sizeOrientation neededRatio format: 1:1, 16:9, 9:16, 2:3, 3:2, 4:3, 3:4 etc.
nWants variations1–4 images
image_urlsEdit/reference imagesUp to 14 URLs; triggers i2i mode

Step 5: Generate & Poll

  1. Call generate_image with model: "gemini-3.1-flash-image-preview" → tell user: "Generating with Nano Banana 2 — ~Xs estimated."
  2. Poll check_task every 3–5s. Report progress %.
  3. After 3 consecutive processing: "Still working..."
  4. Completed: Share URLs. "Links expire in 24h — save promptly."
  5. Failed: Show error + suggestion. Offer retry if retryable.
  6. Timeout (5 min): "Taking longer than expected. Task ID: {id} — check again later."

Error Handling

HTTP Errors

ErrorAction
401"API key isn't working. Check at evolink.ai/dashboard/keys"
402"Balance is low. Add credits at evolink.ai/dashboard/billing"
429"Rate limited — wait 30s and retry"
503"Servers busy — retry in a minute"

Task Errors (status: "failed")

CodeRetry?Action
content_policy_violationNoRevise prompt (no celebrities, NSFW, violence)
invalid_parametersNoCheck values against model limits
image_processing_errorNoCheck format/size/URL accessibility
generation_timeoutYesRetry; simplify prompt if repeated
service_errorYesRetry after 1 min
generation_failed_no_contentYesModify prompt, retry

Full error reference: references/image-api-params.md

Without MCP Server

Use Evolink's file hosting API for image uploads (72h expiry). See references/file-api.md for curl commands.

References

  • references/image-api-params.md — Complete API parameters, model details, polling strategy, error codes
  • references/file-api.md — File hosting API (curl upload/list/delete)

如何使用「Nano Banana 2 — AI Image Generation (Gemini 3.1 Flash Image, Google, Evolink)」?

  1. 打开小龙虾AI(Web 或 iOS App)
  2. 点击上方「立即使用」按钮,或在对话框中输入任务描述
  3. 小龙虾AI 会自动匹配并调用「Nano Banana 2 — AI Image Generation (Gemini 3.1 Flash Image, Google, Evolink)」技能完成任务
  4. 结果即时呈现,支持继续对话优化

相关技能