Configuring Strapi Media Library: S3, Cloudinary, R2

You deployed Strapi to production, and a month later files are lost during deployment? Or the backend crashes under load when serving static assets? A familiar situation: the standard Strapi media library stores everything in public/uploads — fine for development, but critical for production. In this article, we'll walk through configuring Strapi storage for production. Over 10 years, we have configured media libraries for 50+ Strapi projects and know how to avoid these issues. We integrate cloud storage based on S3, Cloudinary, or R2, add automatic image optimization, and file type control. We'll cover a proper configuration that eliminates headaches.

Local storage does not solve three key problems. First, files are not replicated across instances: with horizontal scaling, each server sees only its own files. Second, local files are lost when containers are rebuilt (Docker recreates the layer). And finally, performance: serving static files through Node.js is slower than through a cloud provider's CDN. A correct media library configuration reduces page load time by 30–40% and eliminates backup risks. In one month, local storage can lead to significant losses due to downtime, and image optimization saved one client a substantial amount on traffic.

Why You Shouldn't Use Local Storage in Production

Local storage in public/uploads is only suitable for development. In production, it creates bottlenecks: no fault tolerance, difficult scaling, and files can be lost with every deployment. Cloud storage solves all these problems: files are stored centrally, accessible from anywhere via CDN, and are not lost when containers restart.

Which Provider to Choose for the Strapi Media Library?

Cloudinary processes images 2-3 times faster than S3 with Lambda functions. R2 wins on cost with high traffic volumes — no outbound bandwidth charges.

Configuring AWS S3

Install the package: npm install @strapi/provider-upload-aws-s3. Configuration in config/plugins.js:

module.exports = ({ env }) => ({
  upload: {
    config: {
      provider: 'aws-s3',
      providerOptions: {
        s3Options: {
          credentials: {
            accessKeyId: env('AWS_ACCESS_KEY_ID'),
            secretAccessKey: env('AWS_ACCESS_KEY_SECRET'),
          },
          region: env('AWS_REGION', 'eu-central-1'),
          endpoint: env('AWS_ENDPOINT', undefined), // for S3-compatible
          params: {
            ACL: env('AWS_ACL', 'public-read'),
            Bucket: env('AWS_BUCKET'),
          },
        },
      },
      actionOptions: {
        upload: {},
        uploadStream: {},
        delete: {},
      },
    },
  },
});

Amazon S3 is an object storage service from AWS, widely used for file storage.

Configuring Cloudflare R2

R2 is S3 API-compatible, with free egress:

module.exports = ({ env }) => ({
  upload: {
    config: {
      provider: 'aws-s3',
      providerOptions: {
        s3Options: {
          credentials: {
            accessKeyId: env('R2_ACCESS_KEY_ID'),
            secretAccessKey: env('R2_SECRET_ACCESS_KEY'),
          },
          region: 'auto',
          endpoint: `https://${env('CF_ACCOUNT_ID')}.r2.cloudflarestorage.com`,
          params: {
            Bucket: env('R2_BUCKET'),
          },
        },
      },
    },
  },
});

For public file access, you need a custom R2 domain or a Cloudflare Worker.

Configuring Cloudinary

Install the package: npm install @strapi/provider-upload-cloudinary. Configuration:

module.exports = ({ env }) => ({
  upload: {
    config: {
      provider: 'cloudinary',
      providerOptions: {
        cloud_name: env('CLOUDINARY_NAME'),
        api_key: env('CLOUDINARY_KEY'),
        api_secret: env('CLOUDINARY_SECRET'),
      },
      actionOptions: {
        upload: {
          folder: env('CLOUDINARY_FOLDER', 'strapi'),
          transformation: [{ quality: 'auto', fetch_format: 'auto' }],
        },
        uploadStream: {},
        delete: {},
      },
    },
  },
});

In one project with 500,000 images, we migrated the media library to Cloudinary in 2 days, reducing page load time from 4.2s to 1.1s. This was made possible by automatic optimization and CDN.

Image Formats and Breakpoints

Strapi automatically creates multiple sizes on upload. Configuration in config/plugins.js:

breakpoints: {
  xlarge: 1920,
  large:  1000,
  medium: 750,
  small:  500,
  xsmall: 64,
},

In the API response, the formats field contains URLs for each variant. Sizes and typical usage:

File Type Filtering and Size Limits

Limiting upload size and MIME types:

// config/plugins.js
module.exports = {
  upload: {
    config: {
      sizeLimit: 20 * 1024 * 1024, // 20 MB
    },
  },
};

// src/extensions/upload/strapi-server.js
module.exports = (plugin) => {
  const originalUpload = plugin.controllers.upload.upload;

  plugin.controllers.upload.upload = async (ctx) => {
    const { files } = ctx.request;
    const allowed = ['image/jpeg', 'image/png', 'image/webp', 'application/pdf'];

    const invalid = Object.values(files).flat()
      .filter(f => !allowed.includes(f.type));

    if (invalid.length) {
      return ctx.badRequest('Unsupported file type');
    }

    return originalUpload(ctx);
  };

  return plugin;
};

Configuration Process

1. Requirements analysis: determine file volume, budget, performance needs.
2. Provider selection: based on cost, user geography, transformation needs.
3. Provider configuration: install package, set environment variables, test connection.
4. Breakpoint configuration: adjust sizes to match site design (typically 640, 1024, 1920).
5. Filtering and limits: configure allowed MIME types and maximum file size.
6. Data migration: transfer existing files from local disk and update database references.
7. Testing: verify upload, display, and performance.
8. Documentation and training: hand over instructions to the client's team.

Timelines

• Setting up S3/R2/Cloudinary provider — a few hours.
• Migrating existing files from local storage — 1 day (depends on volume).
• Custom type filtering + limits — a few hours.

What's Included

• Provider storage configuration (S3, Cloudinary, R2).
• Breakpoint configuration and image optimization.
• File type and size filtering implementation.
• Migration of existing files to cloud storage.
• Documentation and instructions for your team.
• Guaranteed correct operation and support for one month after setup.

Get a consultation on configuring your Strapi media library — we'll help you choose the optimal solution and implement it without downtime. Contact us to discuss your project. Order an audit of your current configuration — we'll check for bottlenecks.

Headless CMS: Strapi, Directus, Sanity, Contentful, Drupal

Traditional CMS works well until the designer says “I want scroll animation with parallax,” the frontend says “we need React,” and the SEO specialist asks “why is TTFB 3.4 seconds?” At that point, monolithic architecture starts to hinder everyone. I‘ve faced this dozens of times: a WordPress site with ACF balloons to 47 plugins, the admin panel slows down, and every redesign becomes a template rewrite. Headless CMS separates content management from presentation. Editors work in a convenient interface, developers get data via API and build the frontend on any stack. Sounds simple. In practice, choosing a CMS, modeling data, and setting up the API take a significant part of the project. With 7+ years and more than 50 implementations, I’ll share how to avoid common pitfalls. Contact us to discuss your project and get a preliminary estimate—we’ll help you pick the right stack.

What are the key benefits of headless CMS over monolithic?

Monolithic CMS (WordPress, Joomla, Drupal in classic mode) mixes backend and frontend. Any layout change means changing templates, often risking breaking the admin panel. Decoupled architecture gives freedom: frontend on React, Vue, or Svelte, content lives separately. Result: improved load speed (LCP often drops from 4–6 s to 1–1.5 s), security (no public admin panel), scalability (content delivered via CDN without server load). Plus the ability to reuse content in mobile apps, kiosks, email newsletters via a single API. A client we recently helped saw LCP improve from 6.2 s to 1.1 s — a 5.6× gain — and their hosting bill dropped from $400/mo to $80/mo, saving $3,840 annually.

How to choose a headless CMS for your project?

No universal tool exists. The choice depends on team, content complexity, and infrastructure. Let’s break down the key options.

Strapi — open-source, self-hosted, Node.js. Suitable for teams needing data control and API customization. Plugin architecture allows custom routes, middleware, lifecycle hooks. REST and GraphQL out of the box. Deploys in about an hour — three times faster than Drupal. Weakness: versions v4 and v5 are incompatible, migration is painful. Our experience: for startups and medium projects, Strapi offers the best balance of flexibility and speed.

Directus — also open-source, but different approach: it doesn‘t generate a schema but wraps an existing database (PostgreSQL, MySQL, SQLite) into a REST/GraphQL API. If you already have a database, Directus connects without migrations. Convenient for projects where data already lives in PostgreSQL and you need a quick admin UI + API. Saves up to 30% integration time.

Sanity — cloud CMS with real-time editor. Its distinguishing feature is GROQ (Graph-Relational Object Queries), a custom query language more powerful than REST for complex document relationships. Portable Text for structured content. Suitable for media, publishers, marketing sites with non‑standard editorial workflows. Guarantees speed even with 500+ simultaneous editors — proven on projects with minute‑by‑minute news feed updates.

Contentful — enterprise cloud CMS. Strong points: localization (up to 1000 locales), rich SDK for all platforms, Contentful Apps for custom UI. Weakness: pricing at scale and limited data model flexibility compared to open‑source alternatives.

Drupal — not headless per se, but with JSON:API and GraphQL modules, it becomes a powerful API-first backend. Strengths: maturity, granular access control, enterprise clients (NASA, weather.com). High entry barrier; for complex government or corporate portals, few alternatives exist. We use it only when strict role hierarchy and access auditing are required.

Consequences of poor content modeling

Content modeling is critical. Mistakes at this stage are costly. A typical problem: a body field of type rich text for everything. Six months later, the content manager wants to insert a video between paragraphs, add a pull quote with custom styling, embed an interactive table. Rich text can‘t handle that. Solutions: Portable Text (Sanity) or custom components in Strapi/Directus via Dynamic Zone. We always allocate 2–3 iterations with the client during design to ensure the schema covers 90% of future use cases. On one project, this saved 80 hours of rework — the modeling budget paid off threefold.

How we build projects on headless CMS

Frontend for headless CMS almost always uses Next.js (App Router) or Nuxt. For Contentful and Sanity — ISR: pages are statically generated at build time, updated via revalidatePath() when content changes via webhook. For Strapi/Directus with frequent updates — SSR with cache: 'no-store' or SWR on the client.

Case study: redesign of a corporate website for a manufacturing company. Previous site: WordPress with ACF, 200+ pages, 4 languages. Problems: TTFB 3.8 s, editors complained about slow admin. Migrated to Strapi (self-hosted, PostgreSQL), Next.js App Router. Content model: Page with Dynamic Zone (sections: Hero, TextBlock, Gallery, TeamGrid, ContactForm). Localization via Strapi i18n plugin + next-intl on frontend. Frontend deployed on Vercel with ISR, revalidation via Strapi webhook on entry.publish. According to the client: TTFB dropped from 3.8 s to 180 ms (static with CDN) — a 21× improvement. Editors got a clean interface without 47 plugins. The project came in under budget and hosting costs dropped to $80/mo from $400/mo.

Implementation process broken into stages:

1. Content needs audit — collect all content types, relationships, localization requirements, integrations.
2. Data schema design — create models, fields, validation, access roles. Document in Swagger/OpenAPI.
3. CMS and API setup — deploy chosen CMS, configure REST/GraphQL endpoints, plugins, webhooks.
4. Frontend development — connect Next.js/Nuxt, configure ISR/SSR, section components, routing.
5. Content migration (if legacy) — automated loading via API or scripts.
6. Testing — check API endpoints, regression, load testing, Core Web Vitals.
7. Deployment — configure CDN, SSL, CI/CD, monitoring.

How long does implementation take?

The standard path includes all stages. Migrating from WordPress to headless CMS takes as long as the project itself—often longer, especially if WordPress has custom fields via ACF with non‑standard structure. Our typical timelines:

Cost is calculated individually after a brief. Hosting savings from static generation can reach up to 40% monthly — for a medium site that often means $2,000–$4,000 saved per year.

Non‑obvious considerations when choosing a headless CMS

• Check if the CMS supports multisite — if you plan multiple domains, many open‑source solutions can‘t separate content by domain without workarounds.
• Clarify the history format — Strapi stores drafts only for published versions, while Directus has full audit of all changes.
• Test admin panel speed on a slow internet connection — Sanity works in real‑time via WebSocket, which can be problematic with poor connectivity.
• Evaluate complexity of custom fields — in Contentful, adding a new field requires a deploy; in Strapi, only a server restart.
• Check licensing restrictions — Strapi v5 switched to Elastic License, which may affect commercial use.

What is included

• Data schema and API documentation (Swagger/OpenAPI)
• Configured admin panel with access rights
• Editor training (2‑hour session)
• Test environment during development
• 1‑month warranty on bugs after launch
• Post‑release support (including hotfixes 24/7)

Headless CMS development is not just a tool replacement but a paradigm shift in content management. We help make this transition without downtime or data loss. Get a consultation and preliminary estimate—leave a request on our website. Order headless CMS implementation with guaranteed results.