Ad Copy Generation Template

A Mastra template that generates compelling ad copy and promotional images from content provided as plain text, PDF links, or website URLs. Features AI-powered copywriting with web content extraction and image generation capabilities.

Features

  • Content Processing: Extract and analyze content from PDFs, plain text, or website URLs
  • Web Content Extraction: Extract blog posts and articles from websites using AI-powered web browsing
  • Ad Copy Generation: Create headlines, body copy, and CTAs for multiple platforms
  • Image Generation: Generate promotional images using DALL-E 3 via the AI package
  • Platform Optimization: Tailored content for Facebook, Instagram, Google Ads, LinkedIn, and more
  • A/B Testing: Multiple variations for testing and optimization

Quick Start

  1. Install dependencies:

    pnpm install

  2. Set up environment variables: Create a .env file with:

     1OPENAI_API_KEY=your_openai_api_key
 2
 3# For web content extraction (optional - only needed for URL input type)
 4BROWSERBASE_API_KEY=your_browserbase_api_key
 5BROWSERBASE_PROJECT_ID=your_browserbase_project_id
 6
 7# For AWS S3 cloud storage (required for image generation)
 8AWS_REGION=us-east-1
 9AWS_ACCESS_KEY_ID=your_aws_access_key_id
10AWS_SECRET_ACCESS_KEY=your_aws_secret_access_key
11S3_BUCKET_NAME=mastra-generated-images
12S3_PUBLIC_URL_BASE=https://your-bucket-name.s3.amazonaws.com

  3. Run the application:

    pnpm dev

Usage Examples

Using the Workflow

 1import { mastra } from './src/mastra';
 2
 3// Generate ad copy from text content
 4const result = await mastra.runWorkflow('ad-copy-generation-workflow', {
 5  contentInput: 'Your product description or marketing content here...',
 6  inputType: 'text',
 7  platform: 'facebook',
 8  campaignType: 'conversion',
 9  targetAudience: 'Small business owners aged 25-45',
10  tone: 'professional',
11  generateImages: true,
12  imageStyle: 'modern',
13});
14
15// Generate ad copy from PDF
16const pdfResult = await mastra.runWorkflow('ad-copy-generation-workflow', {
17  contentInput: 'https://example.com/product-brochure.pdf',
18  inputType: 'pdf',
19  platform: 'linkedin',
20  campaignType: 'awareness',
21  tone: 'authoritative',
22});
23
24// Generate ad copy from website URL
25const urlResult = await mastra.runWorkflow('ad-copy-generation-workflow', {
26  contentInput: 'https://techcrunch.com/example-startup-article',
27  inputType: 'url',
28  platform: 'instagram',
29  campaignType: 'consideration',
30  targetAudience: 'Tech enthusiasts and early adopters',
31  tone: 'playful',
32  generateImages: true,
33  imageStyle: 'digital_art',
34});

Using Individual Agents

 1// Generate ad copy directly
 2const adCopy = await mastra.getAgent('adCopyAgent').generate([
 3  {
 4    role: 'user',
 5    content: 'Create Facebook ad copy for a new productivity app targeting remote workers',
 6  },
 7]);
 8
 9// Extract content from PDF
10const pdfContent = await mastra.getTool('pdf-content-extractor').execute({
11  context: { pdfUrl: 'https://example.com/whitepaper.pdf' },
12  mastra,
13});

Components

Agents

  1. Ad Copy Agent - Main agent that orchestrates the ad copy generation process
  2. Content Summarizer Agent - Creates marketing-focused summaries from content
  3. Copywriting Agent - Expert copywriter for creating high-converting ad copy
  4. Web Content Agent - Extracts content from websites and blog posts for ad copy generation

Tools

  1. PDF Content Extractor - Downloads and extracts marketing insights from PDFs
  2. Web Content Extractor - Extracts blog posts and articles from websites using AI-powered browsing
  3. Page Navigate Tool - Navigates to web URLs for content extraction
  4. Ad Copy Generator - Creates multiple ad copy variations for different platforms
  5. Image Generator - Generates promotional images using DALL-E 3

Workflows

Ad Copy Generation Workflow: Simplified end-to-end process that:

  1. Extracts content (from PDF, website URL, or text)
  2. Generates optimized ad copy (headline, body, CTA)
  3. Creates a promotional image and uploads to S3

Platform Support

  • Facebook/Instagram: Optimized for social media advertising
  • Google Ads: Search-intent focused copy
  • LinkedIn: Professional, B2B-oriented content
  • Twitter: Concise, trending-aware copy
  • TikTok: Engaging, video-focused copy
  • Generic: Platform-agnostic advertising copy

Campaign Types

  • Awareness: Brand awareness and reach campaigns
  • Consideration: Lead generation and engagement
  • Conversion: Sales and action-driven campaigns
  • Retention: Customer retention and loyalty

Output Examples

The workflow generates:

Ad Copy

  • Single optimized headline
  • Compelling body copy
  • Effective call-to-action

Image

  • One promotional image uploaded to S3
  • Platform-optimized dimensions
  • Style-consistent visual design

Advanced Configuration

Expected Output Structure

 1const result = await mastra.runWorkflow('ad-copy-generation-workflow', {
 2  // ... input params
 3});
 4
 5// Result structure:
 6{
 7  adCopy: {
 8    headline: "Your optimized headline",
 9    body: "Compelling body copy for your ad",
10    cta: "Call to action"
11  },
12  imageUrl: "https://your-bucket.s3.amazonaws.com/generated-images/uuid.jpg" // Optional
13}

Specific Focus Areas

 1const result = await mastra.getTool('pdf-content-extractor').execute({
 2  context: {
 3    pdfUrl: 'https://example.com/content.pdf',
 4    focusAreas: ['benefits', 'pricing', 'testimonials'],
 5  },
 6  mastra,
 7});

Development

To modify or extend this template:

  1. Add new platforms: Update the platform enums in the schemas
  2. Customize copy styles: Modify the copywriting agent instructions
  3. Add new image styles: Extend the image generation tool
  4. Create new workflows: Combine tools and agents in different ways

Environment Variables

Required

  • OPENAI_API_KEY: Required for AI generation (OpenAI GPT-4)

Optional (based on features used)

  • BROWSERBASE_API_KEY: Required for web content extraction (BrowserBase)
  • BROWSERBASE_PROJECT_ID: Required for web content extraction (BrowserBase)

AWS S3 Cloud Storage (required for image generation)

  • AWS_REGION: AWS region (default: 'us-east-1')
  • AWS_ACCESS_KEY_ID: AWS access key ID
  • AWS_SECRET_ACCESS_KEY: AWS secret access key
  • S3_BUCKET_NAME: S3 bucket name for storing generated images (default: 'mastra-generated-images')
  • S3_PUBLIC_URL_BASE: Public URL base for accessing uploaded images

AWS S3 Setup Example

 1AWS_REGION=us-east-1
 2AWS_ACCESS_KEY_ID=your_aws_access_key_id
 3AWS_SECRET_ACCESS_KEY=your_aws_secret_access_key
 4S3_BUCKET_NAME=mastra-generated-images
 5S3_PUBLIC_URL_BASE=https://mastra-generated-images.s3.amazonaws.com

Note: Make sure your S3 bucket is configured for public read access for the generated images to be accessible via the public URLs.

Dependencies

  • @mastra/core: Core Mastra framework
  • @ai-sdk/openai: OpenAI integration
  • @aws-sdk/client-s3: S3-compatible cloud storage (for image uploads)
  • @browserbasehq/stagehand: Web browsing and content extraction
  • pdf2json: PDF text extraction
  • ai: AI SDK for image generation
  • zod: Schema validation