Google AI Studio (Gemini API)

Access to Google's Gemini models through AI Studio API for text generation, image generation (Nano Banana Pro), embeddings, and multimodal AI capabilities.

Quick Start

bash

# Test authentication
deno run --allow-net --allow-env --allow-read \
  .claude/skills/google/ai-studio/scripts/ai-studio-client.ts test-auth

# Generate an image with Nano Banana Pro
deno run --allow-net --allow-env --allow-read --allow-write \
  .claude/skills/google/ai-studio/scripts/ai-studio-client.ts generate-image \
  "A professional product photo of casual shorts on a beach" \
  --output assets/product-shot.png

# Generate text
deno run --allow-net --allow-env --allow-read \
  .claude/skills/google/ai-studio/scripts/ai-studio-client.ts generate \
  "Write a tagline for a beach clothing brand"

Authentication

Unlike other Google services, AI Studio uses a simple API key (not OAuth2).

Setup

•Go to Google AI Studio
•Click "Get API key" or go to API keys section
•Create a new API key
•Add to .env:

bash

GEMINI_API_KEY="your-api-key-here"

Test Authentication

bash

deno run --allow-net --allow-env --allow-read \
  .claude/skills/google/ai-studio/scripts/ai-studio-client.ts test-auth

Available Commands

Authentication & Models

Command	Description
`test-auth`	Verify API key is valid
`list-models`	List available Gemini models

Text Generation

Command	Description
`generate "<prompt>" [options]`	Generate text content
`stream "<prompt>" [options]`	Stream text generation (real-time output)
`count-tokens "<prompt>" [--model]`	Count tokens in a prompt

Options for generate/stream:

•--model MODEL - Model to use (default: gemini-2.0-flash)
•--temperature TEMP - Temperature 0.0-2.0 (default: 1.0)
•--max-tokens N - Maximum output tokens
•--system "INSTRUCTION" - System instruction

Image Generation (Nano Banana Pro)

Command	Description
`generate-image "<prompt>" [--output PATH] [options]`	Generate image
`edit-image "<input>" "<instruction>" --output PATH`	Edit existing image

Options for generate-image:

•--output PATH - Output file path (default: ./generated-<timestamp>.png)
•
--model MODEL - Image model or alias (default: nano-banana-pro)
- •nano-banana-pro or nbp → gemini-3-pro-image-preview (best quality)
- •nano-banana or nb → gemini-2.5-flash-image (faster, better free tier)
•--aspect RATIO - Aspect ratio: 1:1, 16:9, 9:16, 4:3, 3:4
•--references "path1,path2,..." - Reference images for style (up to 14)

Embeddings

Command	Description
`embed "<text>" [--model]`	Generate text embedding
`embed-batch --input FILE --output FILE`	Batch embeddings from file

File Operations

Command	Description
`upload-file <path> [--name NAME]`	Upload file for multimodal use
`list-files`	List uploaded files
`delete-file <file-id>`	Delete an uploaded file
`get-file <file-id>`	Get file metadata

Use Cases

Image Generation with Nano Banana Pro

Generate high-quality images for marketing, product shots, social media:

bash

# Simple image generation
deno run --allow-net --allow-env --allow-read --allow-write \
  .claude/skills/google/ai-studio/scripts/ai-studio-client.ts generate-image \
  "A cute golden retriever puppy playing on a beach at sunset" \
  --output assets/puppy.png

# With specific aspect ratio for Instagram
deno run --allow-net --allow-env --allow-read --allow-write \
  .claude/skills/google/ai-studio/scripts/ai-studio-client.ts generate-image \
  "Minimalist flat lay of summer accessories" \
  --output assets/flatlay.png \
  --aspect 1:1

# With reference images for brand consistency
deno run --allow-net --allow-env --allow-read --allow-write \
  .claude/skills/google/ai-studio/scripts/ai-studio-client.ts generate-image \
  "Product photo of blue shorts in our brand style" \
  --output assets/branded-product.png \
  --references "assets/{workflow-name}/ref1.jpg,assets/{workflow-name}/ref2.jpg"

# Using Nano Banana (faster, better free tier)
deno run --allow-net --allow-env --allow-read --allow-write \
  .claude/skills/google/ai-studio/scripts/ai-studio-client.ts generate-image \
  "A sunny beach scene with palm trees" \
  --output assets/beach.png \
  --model nano-banana

Text Generation

Generate marketing copy, descriptions, summaries:

bash

# Simple generation
deno run --allow-net --allow-env --allow-read \
  .claude/skills/google/ai-studio/scripts/ai-studio-client.ts generate \
  "Write 3 Instagram captions for a beach clothing brand"

# With system instruction
deno run --allow-net --allow-env --allow-read \
  .claude/skills/google/ai-studio/scripts/ai-studio-client.ts generate \
  "Write a product description for casual summer shorts" \
  --system "You are a copywriter for a premium beachwear brand. Write in a casual, friendly tone."

# Streaming for long content
deno run --allow-net --allow-env --allow-read \
  .claude/skills/google/ai-studio/scripts/ai-studio-client.ts stream \
  "Write a blog post about sustainable fashion trends"

Multimodal (Image + Text)

Analyze images by uploading first:

bash

# Upload an image
deno run --allow-net --allow-env --allow-read \
  .claude/skills/google/ai-studio/scripts/ai-studio-client.ts upload-file \
  product-photo.jpg --name "Product Photo"

# Use in generation (returns file URI to use)
deno run --allow-net --allow-env --allow-read \
  .claude/skills/google/ai-studio/scripts/ai-studio-client.ts generate \
  "Describe this product photo and suggest improvements" \
  --file "files/abc123"

Embeddings

Generate embeddings for semantic search:

bash

# Single embedding
deno run --allow-net --allow-env --allow-read \
  .claude/skills/google/ai-studio/scripts/ai-studio-client.ts embed \
  "comfortable beach shorts for summer"

# Batch embeddings
deno run --allow-net --allow-env --allow-read --allow-write \
  .claude/skills/google/ai-studio/scripts/ai-studio-client.ts embed-batch \
  --input products.txt --output embeddings.json

Available Models

Text Generation

•gemini-2.0-flash - Fast, efficient (default)
•gemini-2.0-flash-lite - Fastest, most cost-effective
•gemini-2.5-pro-preview-05-06 - Most capable reasoning
•gemini-2.5-flash-preview-05-20 - Fast with enhanced capabilities

Image Generation

•nano-banana-pro / nbp → gemini-3-pro-image-preview (default, best quality)
•nano-banana / nb → gemini-2.5-flash-image (faster, 500 images/day free tier)

Embeddings

•text-embedding-004 - Latest embedding model (default)

See references/models.md for full model details.

Environment Variables

bash

# Required
GEMINI_API_KEY="your-api-key"

# Optional
GEMINI_DEFAULT_MODEL="gemini-2.0-flash"          # Default text model
GEMINI_IMAGE_MODEL="gemini-3-pro-image-preview"  # Nano Banana Pro
GEMINI_OUTPUT_DIR="."                              # Default image output directory (current working directory)

When to Use This Skill

Use AI Studio when:

•Generating images with Nano Banana Pro
•Generating text with Gemini models
•Creating embeddings for search/similarity
•Processing multimodal content (images + text)
•User mentions "Gemini", "AI Studio", "Nano Banana"

Don't use AI Studio when:

•Working with Google Workspace (Gmail, Drive, Sheets) - use respective google skills
•Need Claude's capabilities - use direct conversation
•Working with other image APIs (Stable Diffusion, DALL-E)

For prompting guidance: See the visual-design skill for comprehensive prompting best practices, templates, and workflows for Your Brand brand imagery.

Troubleshooting

"Invalid API key"

•Verify GEMINI_API_KEY is set in .env
•Check key is valid at Google AI Studio

"Model not found"

•Run list-models to see available models
•Check model name spelling
•Some models may be preview-only

"Quota exceeded"

•Free tier has limited requests/day
•Consider upgrading to paid tier
•Check quota at Google Cloud Console

Image generation fails

•Ensure prompt follows content guidelines
•Try simpler prompts first
•Check that --output path directory exists

Rate Limits

Text Generation (Free Tier):

Model	RPM	RPD
gemini-2.0-flash	15	1,500
gemini-2.5-pro	5	50

Image Generation (Free Tier):

Model	IPM	IPD
gemini-3-pro-image-preview	2	100-500
gemini-2.5-flash-image	2	500

RPM = Requests/min, RPD = Requests/day, IPM = Images/min, IPD = Images/day

Paid Tier: $0.02-0.04 per image for higher limits.

Google AI Studio (Gemini API)

Quick Start

Authentication

Setup

Test Authentication

Available Commands

Authentication & Models

Text Generation

Image Generation (Nano Banana Pro)

Embeddings

File Operations

Use Cases

Image Generation with Nano Banana Pro

Text Generation

Multimodal (Image + Text)

Embeddings

Available Models

Text Generation

Image Generation

Embeddings

Environment Variables

When to Use This Skill

Troubleshooting

"Invalid API key"

"Model not found"

"Quota exceeded"

Image generation fails

Rate Limits

Additional Resources