Free Image API in 2026: Unsplash 1,000/hr, Pexels New Keys Paused

As of October 10, 2026, five free image APIs are worth a developer's time: Unsplash, Pexels, Pixabay, Wikimedia Commons and Openverse. Two changes from 2026 decide which one you can actually start with. Unsplash lowered its production limit from 5,000 to 1,000 requests per hour, and Pexels stopped issuing new API keys in early October with no reopening date. As of November 2, 2026, the Pexels API page still reads "New API key issuance is currently paused and cannot be requested." That message is a notice to every new applicant, not a problem with your account or your country, and Pexels keys that already exist keep working.

If you don't already hold a Pexels key, the short answer is this. Use Unsplash for the best-looking photos in a search UI, as long as you can hotlink its URLs and show credit; 50 requests an hour in demo mode is enough to build, and approval lifts you to 1,000. Use Pixabay if you want to download images and serve them from your own storage. Use Openverse or Wikimedia Commons if you need openly licensed media without a key. For placeholders, Lorem Picsum is simpler than any of these. If you need generated images rather than existing photos, that is a different API category, covered in Free AI Image Generation API in 2026: Daily Image Limits Compared and the paid side in the AI image generation API comparison.

TL;DR: Which free image API should you use?

This chart predates the October 10, 2026 update. Pexels no longer offers fast onboarding to new developers because new key issuance is paused; the text and tables below are current.

Start from what your product does with the images, not from the brand.

If Pexels was your first choice, the closest substitute right now is Pixabay for photos plus video, or Unsplash if photo quality matters more than self-hosting. The deciding question is rarely image quality alone. It is whether your interface can show attribution, whether you are allowed to hotlink or must store files yourself, whether you need a key today, and whether your team can review licenses file by file.

Free image search API rate limits and rules as of October 10, 2026

These are the published limits and key status for each free image search API, taken from each provider's own documentation. The Pexels and Pixabay key status and default rate limits were rechecked on November 2, 2026.

Three of those entries changed in 2026. The 5,000-per-hour Unsplash figure that still circulates is the pre-change number.

• Unsplash: 5,000 → 1,000 per hour in production. The Unsplash API documentation said 5,000 in an Internet Archive snapshot from March 26, 2026 and has said 1,000 since at least May 2, 2026. Its header example now shows X-Ratelimit-Limit: 1000.
• Pexels: new keys paused. The Pexels API page now states "New API key issuance is currently paused and cannot be requested." Internet Archive snapshots place the notice between September 28 and October 4, 2026, and it was still there on November 2, 2026. Pexels gives no reason and no end date.
• Wikimedia: published per-minute limits. The Wikimedia API rate limits page now lists numbers by client type. Enforcement began in phases on March 12, 2026 and the week of April 28, 2026, and the page says the limits are "subject to experimentation and change".

What people actually mean when they say “free image API”

The phrase covers several different tools, and each has its own cost model and legal rules:

• Stock-photo search APIs return existing photos uploaded by photographers or institutions. Unsplash, Pexels, Pixabay, Wikimedia Commons and Openverse belong here.
• Placeholder services such as Lorem Picsum hand you a random or seeded image by URL shape (/200/300, /seed/{seed}/200/300). They are for mockups, not for content.
• AI image generation APIs create new images from a prompt. The questions there are free daily quota, price per image and moderation, not attribution. Current free quotas are compared in Free AI Image Generation API in 2026: Daily Image Limits Compared.
• Image upload and hosting APIs store your own files and serve them. That is a storage or image-CDN job; none of the stock-photo APIs here are meant to host your uploads.
• Google Images. Google has no official free API for its image search. Services such as SerpApi's Google Images API scrape the results page; SerpApi's free plan allows 250 searches a month (50 an hour), and the results are search hits, not images licensed for reuse.

The phrase also hides a legal trap. A site can describe its photos as free to use, and that can be true for downloads from its website, while its API still adds rules about attribution, hotlinking, download tracking and product design. Unsplash and Pexels both work this way. That is why the API documentation, not the license summary, should drive your implementation.

Unsplash vs Pexels vs Pixabay vs Wikimedia Commons

All four are free to call, but each makes your app do something different: Unsplash wants hotlinks and credit, Pexels can't issue new keys right now, Pixabay wants downloads and caching, and Wikimedia wants a User-Agent and per-file license checks.

Unsplash: 50 requests/hour in demo, 1,000/hour in production

Unsplash is the best default when photo quality matters more than implementation freedom. New applications start in demo mode at 50 requests per hour. After Unsplash approves the app for production, the limit rises to 1,000 per hour. Only JSON requests to api.unsplash.com count against it; loading the image files themselves does not. Registration for new developers is open.

The API comes with a compliance model. The Unsplash API guidelines require you to:

• hotlink the image URLs from photo.urls, for every use of the image, not only in search results;
• call photo.links.download_location whenever your app does something download-like;
• credit Unsplash and the photographer with links that carry ?utm_source=your_app_name&utm_medium=referral;
• keep "Unsplash" out of your app's name, and not build a wallpaper app or an unofficial Unsplash client.

Production approval checks those points and asks for screenshots, so build them in from day one. The documentation also warns that "Too many requests too quickly will get your access turned off." The Unsplash API Terms §12 sends anyone using API content for AI or machine learning to Unsplash Datasets, where a Lite dataset of 25,000 images is free and the full dataset is available by contacting Unsplash. The same section lets Unsplash charge a technology fee to certain heavy business users.

The confusing part is that the general Unsplash License allows commercial and non-commercial use with attribution appreciated but not required, while the API rules make attribution and hotlinking mandatory for API integrations. The license also forbids compiling photos to replicate a competing service. Unsplash suits editorial galleries, CMS image pickers, newsletters and magazine-style interfaces where credit lines are natural. It is a poor fit if you want to copy every file into your own storage and hide where it came from.

Pexels: 200/hour and 20,000/month, but new keys are paused

Pexels was long the easiest all-purpose choice: one key, a plain Authorization header, photos and videos from the same API. As of November 2, 2026, the Pexels API page shows "New API key issuance is currently paused and cannot be requested." twice, with no reason and no date for reopening. If you already have a key, nothing changes. If you don't, plan around Unsplash or Pixabay until the notice disappears.

The message sits on the public API page itself, so it applies to everyone asking for a new key. It doesn't mean your account was flagged or that your country is blocked, and there is nothing on your side to fix while the notice is up. What makes it confusing is that the Pexels API documentation still says, as of November 2, 2026, "Anyone with a Pexels account can request an API key, which you will receive instantly." That line is older than the pause notice, which appeared between late September and early October, so the two Pexels pages contradict each other and the API page is the one describing the current state.

If you hit the message today, the substitute depends on what you were going to use Pexels for. For photos and videos from one API, switch to Pixabay and plan for caching and self-hosting. For a search or picker UI where photo quality matters most, use Unsplash, which has open registration and starts you at 50 requests an hour. If you'd rather not register anywhere, Openverse answers anonymous requests without a key. Neither Pexels page says existing keys will stop working, so a project that already has one can stay on Pexels.

For existing keys, the Pexels API documentation sets defaults of 200 requests per hour and 20,000 per month. Pexels states that "If you meet our API terms, you can get unlimited requests for free." Successful (2xx) responses include X-Ratelimit-Limit, X-Ratelimit-Remaining and X-Ratelimit-Reset headers, with the reset tied to the monthly quota, so read them from a successful call rather than from an error.

The Pexels License says photos and videos are free to use and attribution is not required, but the API docs still ask you to show a prominent link to Pexels and to credit photographers when possible. The Pexels Terms of Service §8 also bans scraping "including without limitation for machine learning purposes", bulk or systematic copying, and compiling content into a competing service.

Pixabay: 100 requests per 60 seconds, cached for 24 hours

Pixabay is the practical answer when you want to download and serve images from your own infrastructure. The Pixabay API documentation shows your key once you are logged in. It allows up to 100 requests per 60 seconds per key and returns at most 500 images per query. A Reddit thread from late October 2026 claims that Pixabay, like Pexels, has stopped issuing new keys, but as of November 2, 2026 the Pixabay API documentation shows no pause notice and still says "Please login to see your API key here", so treat the claim as unconfirmed. If no key shows up on that page after you log in, that is the first sign the report is right.

Its rules point you toward a download-first design:

• requests must be cached for 24 hours, and you shouldn't send lots of automated queries; systematic mass downloads are not allowed;
• permanent hotlinking of images is not allowed, so download the images you use to your own server first; videos may be embedded directly;
• the webformatURL images (up to 640 px) are only valid for 24 hours, which is another reason to store what you keep;
• full-resolution imageURL and vectorURL fields require approval from Pixabay;
• wherever you show search results, show users where the images and videos come from.

Pixabay covers photos, illustrations, vectors and videos, which suits CMS importers, marketing tools and blog pipelines. Its Content License Summary allows free use without attribution and with modification, but restricts standalone resale, misleading use, and some cases involving trademarks or recognizable people. The Pixabay terms §8 forbids scraping for machine-learning purposes.

Wikimedia Commons: no key, but a User-Agent and per-minute limits

Wikimedia Commons is the power-user choice, not the default. It is the least polished as a drop-in stock-photo service, but the most useful when you need open media with file-level metadata and provenance. No key is needed. The Wikimedia User-Agent policy does require a meaningful User-Agent with contact information, and requests without one can be refused with HTTP 403.

Since 2026, the Wikimedia API rate limits page publishes per-minute numbers:

Keep three or fewer requests in flight. When you receive HTTP 429, honor the Retry-After header, and if it is missing, wait at least 5 seconds. The separate Wikimedia robot policy is stricter for unauthenticated Action API clients: one request at a time and below 5 requests per second, with media downloads at no more than two concurrent connections and 25 Mbps. The two pages don't use the same figures, so staying within both (one request at a time, under 200 a minute) is the safe reading.

Licensing is per file. The Commons reuse guidance lists CC BY, CC BY-SA, GFDL and public domain among the licenses you will meet, asks you to credit the author and the license, and warns that "Other restrictions may apply". Hotlinking Commons files is possible but not recommended. A new Attribution API, in beta, is meant to make credit lines easier to build. High-volume commercial users are directed to Wikimedia Enterprise or the database dumps rather than the live API. Commons fits educational tools, archives, cultural projects and knowledge products; for a SaaS dashboard that just needs pleasant background photos, it is usually too much work.

Openverse, Lorem Picsum and Flickr: keyless and niche options

Openverse is a search API over openly licensed and public-domain media, and it works without a key. Anonymous responses carry rate-limit headers of 20 requests per minute (burst) and 200 per day (sustained). The Openverse documentation gives no numbers, but registering an OAuth2 application raises the limits slightly. Each result includes license fields and a ready-made attribution string, and the license_type=commercial filter narrows results to licenses that allow commercial use. Its terms forbid scraping and using multiple machines to get around the limits. Openverse is the closest answer to "free images API without a key" for production use, as long as 200 requests a day fits your traffic or you cache results.

Lorem Picsum serves placeholder images straight from URLs: /200/300 for a random image, /seed/{seed}/200/300 for a repeatable one, plus ?grayscale and ?blur=. The images come from Unsplash. The site publishes no rate limits and no terms page, which is fine for mockups and prototypes but not a base for production features.

Flickr still offers free API keys for non-commercial use. Commercial apps must apply separately, and "fees may apply". The Flickr API terms require the notice "This product uses the Flickr API but is not endorsed or certified by SmugMug, Inc." and limit display to 30 Flickr user photos per page.

The rules that make or break a free image API integration

A content license is not the whole contract. The license tells you how the image may be reused. The API rules tell you how your application must behave, and breaking them is what costs you access: Unsplash, for one, turns off clients that send too many requests too quickly.

Attribution and linkback. If your interface has a natural place for "Photo by X on Unsplash" or "Photos provided by Pexels", those two services are easy to adopt. If it has nowhere to show credit, they are harder than they look. Unsplash's guidelines require photographer and Unsplash credit with referral links; Pexels asks for a prominent link to Pexels; Pixabay asks you to show where search results come from; Wikimedia and Openverse need the author and license for each file.

Hotlinking versus self-hosting. Unsplash requires you to hotlink its URLs for every use. Pixabay forbids permanent hotlinking of images and makes its preview URLs expire after 24 hours, so you must download what you keep. This decides whether your image layer lives in the frontend, in a proxy, or in your media pipeline, so settle it before you write storage code.

Rate-limit behavior, not just rate-limit numbers. Unsplash turns off access for "too many requests too quickly", Pixabay forbids lots of automated queries and requires caching, and Wikimedia returns 429 with Retry-After. Cache search results, back off on 429, and keep request bursts small. A quota that looks generous on paper doesn't protect a client that hammers the API.

AI and machine-learning use. Pexels and Pixabay terms ban scraping for machine learning. Unsplash routes AI/ML use to its datasets instead of the API. Openverse forbids scraping. If you plan to train or fine-tune a model, none of these APIs is the right channel.

"Free for commercial use" is not the same as "safe for every commercial product". Pixabay's license summary restricts some uses involving brands, logos and recognizable people. Wikimedia Commons warns that rights beyond copyright may apply and that file-level license information is not guaranteed. Openverse results can include non-commercial licenses such as CC BY-NC, and Commons results mix public-domain files with share-alike ones, so one result set can carry very different obligations. If your product shows media to users at scale, check these caveats before launch.

A high request quota doesn't help if the API expects a presentation model your product can't support. Pick the API whose rules match your delivery model first, then check whether its limit covers your traffic.

Fastest working integration path for each API

Each snippet below is a first request in the format the provider documents, followed by what production use requires next. Keys come from environment variables; replace the User-Agent contact details with your own.

Unsplash: search, render the returned URLs, and build attribution into the UI from day one.

curl -H "Authorization: Client-ID $UNSPLASH_ACCESS_KEY" \
  "https://api.unsplash.com/search/photos?query=workspace&per_page=5"

Render the URLs from photo.urls directly and keep their parameters. When your user downloads or selects an image for use, call the photo.links.download_location URL from the response. Add the credit line with the utm_source and utm_medium parameters before you apply for production, because approval reviews those points.

Pexels: only if you already have a key.

curl -H "Authorization: $PEXELS_API_KEY" \
  "https://api.pexels.com/v1/search?query=workspace&per_page=5"

The src object in each result holds several sizes, so pick the one that matches your layout. Place the prominent Pexels link next to the results, and watch the X-Ratelimit-Remaining header to see how much of the monthly quota is left.

Pixabay: treat it as an ingestion source, not a permanent image host.

curl "https://pixabay.com/api/?key=$PIXABAY_KEY&q=workspace&image_type=photo&safesearch=true&per_page=20"

Cache each query for 24 hours, download the images you decide to use, and serve them from your own storage or CDN. The webformatURL links expire after 24 hours, so don't save them in your database as permanent image URLs.

Wikimedia Commons: send a User-Agent and read the license metadata first.

curl -A "YourApp/1.0 (https://example.com; you@example.com)" \
  "https://commons.wikimedia.org/w/api.php?action=query&generator=search&gsrsearch=workspace&gsrnamespace=6&gsrlimit=5&prop=imageinfo&iiprop=url|extmetadata&format=json"

gsrnamespace=6 limits the search to files, and extmetadata returns the author, license and credit fields. Check those before you show or store an image. Send requests one at a time and back off on 429.

Openverse: no key, license filter built in.

curl "https://api.openverse.org/v1/images/?q=workspace&license_type=commercial&page_size=5"

Use the attribution field from each result as your credit line, and cache results so you stay under the anonymous limit of 200 requests a day.

The first request is the easy part. Matching your delivery model to the provider's rules is what saves or costs you a rewrite.

So which free image API should you start with?

As of October 10, 2026, the honest starting point depends on whether you can get a key today and how you deliver images.

Start with Unsplash if you are building a search or picker UI and photo quality matters. Registration is open, demo mode gives you 50 requests an hour to build with, and production approval raises that to 1,000. Treat hotlinking, download tracking and credit as product requirements, not afterthoughts.

Start with Pixabay if you want operational control: ingest assets, resize them in your own pipeline, and serve them from your own domain. It is the best fit for self-hosting and the closest substitute for Pexels when you need photos and videos from one API.

Keep using Pexels if you already have a key. It still offers 200 requests an hour and 20,000 a month by default, and unlimited free requests if you meet its terms. Don't plan a new project around it until new keys are available again.

Choose Openverse or Wikimedia Commons when you need no key, open licenses and provenance more than a curated stock feed. Openverse is quicker to integrate; Commons gives deeper metadata but needs a User-Agent and careful per-file license checks.

There is no universal winner. Pick the API whose rules match your product, and recheck the limits when you scale, because three of these five services changed their access rules in 2026.

FAQ

Which free image API is best for commercial use?

For most commercial apps as of October 10, 2026, Unsplash or Pixabay. Unsplash suits apps that can hotlink and show credit; Pixabay suits apps that download and self-host images. Pexels is equally good for commercial use if you already have a key, but new keys are paused. With Wikimedia Commons and Openverse, commercial use depends on each file's license; Openverse's license_type=commercial filter helps.

Is Unsplash really free if the API requires attribution?

Yes, but "free" doesn't mean "no obligations". The Unsplash License lets you use photos commercially without required attribution, yet the API guidelines require attribution, hotlinking and download tracking for apps built on the API. Separate the content-license question from the API-integration question.

Can I still get a Pexels API key?

Not as of November 2, 2026. The Pexels API page shows "New API key issuance is currently paused and cannot be requested." without a reason or reopening date, and the pause applies to all new applicants, not to particular accounts or countries. The Pexels documentation page still says new keys arrive instantly, but that text predates the pause. Existing keys keep working with the default 200 requests per hour and 20,000 per month. For a new project today, Pixabay is the closest substitute for photos and videos, and Unsplash is the better pick for photo quality.

Is there a free image API without a key?

Yes. Openverse works anonymously with about 20 requests per minute and 200 per day, and returns license and attribution data. Wikimedia Commons needs no key but requires a User-Agent with contact details, and allows 200 requests per minute for identified clients. Lorem Picsum needs no key for placeholder images.

Is the Google Images API free?

Google has no official free API for searching Google Images. Third-party services such as SerpApi scrape Google's results; SerpApi's free plan allows 250 searches a month. Those results are search hits, not images you are licensed to reuse.

Can I store images on my own server?

Pixabay is built for that: download the images you use, cache requests for 24 hours, and don't hotlink images permanently. Unsplash is the least suitable, because its API requires hotlinking the returned URLs for every use. Wikimedia Commons and Openverse allow downloading and reuse under each file's license.

Which free image API supports videos too?

Pixabay and Pexels both return videos through their APIs; Pixabay also allows videos to be embedded directly. Unsplash's API is built around photos. Wikimedia Commons hosts video files too, but with the same per-file licensing as its images.

Do I have to check licenses file by file on Wikimedia Commons?

Yes. Commons files carry different licenses, including CC BY, CC BY-SA, GFDL and public domain, each with its own attribution or share-alike terms. Read the extmetadata fields for every file you use. If you want one uniform license, Unsplash or Pixabay is simpler.

Can I use free image API results to train an AI model?

Not through these APIs. Pexels and Pixabay terms ban scraping for machine-learning purposes, Openverse forbids scraping, and Unsplash's API Terms point AI and ML use to Unsplash Datasets, where a 25,000-image Lite dataset is free.