Image Content Moderation

Screen images for adult, violent, racy, medical, and spoof content using vision-capable LLMs. Returns confidence scores for each category with token usage tracking.

What it does

Analyzes images for inappropriate content across five categories:

• Adult - Explicit adult content
• Violence - Violent or graphic content
• Racy - Suggestive but not explicit content
• Medical - Medical or graphic imagery
• Spoof - Memes, edited, or manipulated images

Returns a boolean safe verdict plus confidence scores (0-1) for each category.

Input

Provide an image URL or base64 data. API keys are set via environment variables.

{
  "imageUrl": "https://example.com/photo.jpg",
  "model": "openai:gpt-4o"
}

Or with base64:

{
  "imageBase64": "iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAADUlEQVR42mNk...",
  "model": "openai:gpt-4o"
}

Model Selection

Use provider:model format to specify any vision-capable model:

{
  "model": "openai:gpt-4o"
}

Supported providers and example models:

Provider	Example Models	Environment Variable
openai	openai:gpt-4o, openai:gpt-4o-mini, openai:gpt-4-turbo	OPENAI_API_KEY
anthropic	anthropic:claude-sonnet-4-20250514, anthropic:claude-3-5-sonnet-20241022	ANTHROPIC_API_KEY
google	google:gemini-2.0-flash, google:gemini-1.5-pro	GOOGLE_API_KEY
groq	groq:meta-llama/llama-4-scout-17b-16e-instruct	GROQ_API_KEY

Default: openai:gpt-4o

Custom Thresholds

Override thresholds per category (0-1 scale, higher = stricter):

{
  "imageUrl": "https://example.com/photo.jpg",
  "model": "openai:gpt-4o-mini",
  "config": {
    "adult": 0.3,
    "violence": 0.5,
    "racy": 0.7
  }
}

Defaults:

• adult: 1 (very strict)
• violence: 1 (very strict)
• racy: 1 (very strict)
• medical: 0.5
• spoof: 0.5

Set to 0 to skip that category.

Output

{
  "safe": true,
  "flagged": [],
  "scores": {
    "adult": 0.1,
    "violence": 0.0,
    "racy": 0.3,
    "medical": 0.0,
    "spoof": 0.1
  },
  "configApplied": {
    "adult": 1,
    "violence": 1,
    "racy": 1,
    "medical": 0.5,
    "spoof": 0.5
  },
  "imageUrl": "https://example.com/photo.jpg",
  "processedAt": "2024-12-21T10:30:00.000Z",
  "usage": {
    "tokens": { "input": 1250, "output": 85, "total": 1335 },
    "cost": 0.003825
  }
}

When content is flagged:

{
  "safe": false,
  "flagged": ["adult", "racy"],
  "scores": {
    "adult": 0.9,
    "violence": 0.0,
    "racy": 0.7,
    "medical": 0.0,
    "spoof": 0.1
  }
}

API Keys

Set API keys as environment variables:

Provider	Environment Variable
openai:*	OPENAI_API_KEY
anthropic:*	ANTHROPIC_API_KEY
google:*	GOOGLE_API_KEY
groq:*	GROQ_API_KEY

Get your API keys:

• OpenAI: platform.openai.com/api-keys
• Anthropic: console.anthropic.com
• Google: aistudio.google.com/apikey
• Groq: console.groq.com

Use Cases

• User-Generated Content - Filter uploads before publishing
• E-commerce - Validate product images
• Social Platforms - Automated content screening
• Ad Networks - Ensure ad creative compliance
• Media Libraries - Bulk image classification

API Reference

Input

Field	Type	Required	Description
imageUrl	string	One of	Public URL of image to analyze
imageBase64	string	One of	Base64-encoded image data
model	string	No	LLM model in provider:model format (default: openai:gpt-4o)
config	object	No	Override thresholds per category (0-1)

Output

Field	Type	Description
safe	boolean	true if no categories exceed thresholds
flagged	string[]	Categories that exceeded thresholds
scores	object	Confidence scores (0-1) for each category
configApplied	object	Config used (defaults + your overrides)
imageUrl	string	Original image URL (if provided)
processedAt	string	ISO timestamp
usage.tokens	object	Token counts (input, output, total)
usage.cost	number	Estimated cost in USD

Limits

• Images must be under 20MB
• Supported formats: JPEG, PNG, GIF, BMP, WEBP