The journey from manually creating OpenGraph images to implementing an automated API-driven system represents a critical evolution for growing web applications. Today, I'll share how I transformed this process at gleam.so, moving from individual Figma designs to an automated system handling thousands of images.
The Manual Phase: Understanding the Baseline
Initially, like many developers, I created OG images manually:
// Early implementation
const getOGImage = (postId: string) => {
return `/images/og/${postId}.png`; // Manually created in Figma
};
This process typically involved:
- Opening Figma for each new image
- Adjusting text and elements
- Exporting to the correct size
- Uploading and linking the image
Average time per image: 15-20 minutes.
First Step: Templating System
The first automation step involved creating reusable templates:
interface OGTemplate {
layout: string;
styles: {
title: TextStyle;
description?: TextStyle;
background: BackgroundStyle;
};
dimensions: {
width: number;
height: number;
};
}
const generateFromTemplate = async (
template: OGTemplate,
content: Content
): Promise<Buffer> => {
const svg = renderTemplate(template, content);
return convertToImage(svg);
};
This reduced creation time to 5 minutes per image but still required manual intervention.
Building the API Layer
The next evolution introduced a proper API:
// api/og/route.ts
import { ImageResponse } from '@vercel/og';
import { getTemplate } from '@/lib/templates';
export const config = {
runtime: 'edge',
};
export async function GET(request: Request) {
try {
const { searchParams } = new URL(request.url);
const template = getTemplate(searchParams.get('template') || 'default');
const content = {
title: searchParams.get('title'),
description: searchParams.get('description'),
};
const imageResponse = new ImageResponse(
renderTemplate(template, content),
{
width: 1200,
height: 630,
}
);
return imageResponse;
} catch (error) {
console.error('OG Generation failed:', error);
return new Response('Failed to generate image', { status: 500 });
}
}
Implementing Caching Layers
Performance optimization required multiple caching layers:
class OGCache {
private readonly memory = new Map<string, Buffer>();
private readonly redis: Redis;
private readonly cdn: CDNStorage;
async getImage(key: string): Promise<Buffer | null> {
// Memory cache
if (this.memory.has(key)) {
return this.memory.get(key);
}
// Redis cache
const redisResult = await this.redis.get(key);
if (redisResult) {
this.memory.set(key, redisResult);
return redisResult;
}
// CDN cache
const cdnResult = await this.cdn.get(key);
if (cdnResult) {
await this.warmCache(key, cdnResult);
return cdnResult;
}
return null;
}
}
Resource Optimization
Handling increased load required careful resource management:
class ResourceManager {
private readonly queue: Queue;
private readonly maxConcurrent = 50;
private activeJobs = 0;
async processRequest(params: GenerationParams): Promise<Buffer> {
if (this.activeJobs >= this.maxConcurrent) {
return this.queue.add(params);
}
this.activeJobs++;
try {
return await this.generateImage(params);
} finally {
this.activeJobs--;
}
}
}
Integration Example
Here's how it all comes together in a Next.js application:
// components/OGImage.tsx
export function OGImage({ title, description, template = 'default' }) {
const ogUrl = useMemo(() => {
const params = new URLSearchParams({
title,
description,
template,
});
return `/api/og?${params.toString()}`;
}, [title, description, template]);
return (
<Head>
<meta property="og:image" content={ogUrl} />
<meta property="og:image:width" content="1200" />
<meta property="og:image:height" content="630" />
</Head>
);
}
Performance Results
The automated system achieved significant improvements:
- Generation time: <100ms (down from 15-20 minutes)
- Cache hit rate: 95%
- Error rate: <0.1%
- CPU usage: 15% of previous implementation
- Cost per image: $0.0001 (down from ~$5 in manual labor)
Key Learnings
Through this automation journey, several crucial insights emerged:
-
Image Generation Strategy
- Pre-warm caches for predictable content
- Implement fallbacks for failures
- Optimize template rendering first
-
Resource Management
- Implement request queuing
- Monitor memory usage
- Cache aggressively
-
Error Handling
- Provide fallback images
- Log failures comprehensively
- Monitor generation metrics
The Path Forward
The future of OG image automation lies in:
- AI-enhanced template selection
- Dynamic content optimization
- Predictive cache warming
- Real-time performance tuning
Simplifying Implementation
While building a custom solution offers valuable learning experiences, it requires significant development and maintenance effort. That's why I built gleam.so, which provides this entire automation stack as a service.
Now you can:
- Design templates visually
- Preview all options for free
- Generate images via API (Open beta-test for lifetime users)
- Focus on your core product
75% off lifetime access ending soon ✨
Share Your Experience
Have you automated your OG image generation? What challenges did you face? Share your experiences in the comments!
Part of the Making OpenGraph Work series. Follow for more web development insights!
Top comments (0)