Website Screenshot API: full page, mobile, PDF

Give it a list of URLs and it returns one image (PNG, JPEG or WebP) or PDF per URL, stored in the run's key-value store, plus a dataset row with the public link, HTTP status, final URL after redirects, page title, pixel size, file size and SHA-256. A URL that fails produces a row with ok: false and an error message. It never stops the rest of the batch and you are not charged for it.

It is a general-purpose screenshot tool: monitoring pages, link previews and Open Graph images, archiving evidence, QA across devices, thumbnails for a directory, visual input for an AI agent. There is no target site; the URL is yours.

What you get that the basic screenshot actors skip

• Full page that actually works. Scrolls the page in steps first so lazy-loaded images and sections render, then captures. Pages taller than 16,383 device pixels are cut at that limit and flagged with truncated: true.
• Cookie banners removed. Clicks "reject" on the common consent managers (selector lists for OneTrust, Cookiebot, Didomi, Quantcast, Usercentrics, Klaro and others, plus a text match on buttons inside overlays in English, Portuguese, Spanish, French, German and Italian). If a banner only offers "accept", it clicks that. Whatever is left gets hidden and page scrolling is restored. Checked end to end on The Guardian, BBC, Le Monde, Der Spiegel and UOL.
• Sticky headers do not repeat. In full-page mode, position: fixed and sticky elements are pinned to their place instead of showing up in the middle of the image.
• Crop to an element or a rectangle with a CSS selector or {x, y, width, height}.
• Emoji and CJK text render. The image ships with Noto Color Emoji and Noto CJK, so emoji do not turn into empty squares.
• Per-URL errors. DNS failure, TLS error, timeout, HTTP 404, missing selector: each one becomes a row with a readable error.
• Charged only on success. See Pricing.

Quick start

Input in the Console, or via API:

{
  "urls": ["https://www.bbc.com/news"],
  "device": "desktop-1366",
  "fullPage": true,
  "format": "jpeg",
  "quality": 80
}

Output row from a real run. The local run stores files on disk, so screenshotUrl is shown here in the form it takes on the platform:

{
  "url": "https://www.bbc.com/news",
  "finalUrl": "https://www.bbc.com/news",
  "ok": true,
  "httpStatus": 200,
  "error": null,
  "title": "BBC News - Breaking news, video and the latest top stories from the U.S. and around the world",
  "device": "desktop-1366",
  "format": "jpeg",
  "fullPage": true,
  "truncated": false,
  "width": 1366,
  "height": 6601,
  "fileSizeBytes": 1058915,
  "sha256": "caac18061616f8bdedda6dc9d7b553bfc66cdea29cbc76c5f0e31e985153377d",
  "screenshotUrl": "https://api.apify.com/v2/key-value-stores/<storeId>/records/001-www-bbc-com-news.jpg",
  "kvKey": "001-www-bbc-com-news.jpg",
  "loadTimeMs": 1427,
  "totalTimeMs": 5323,
  "capturedAt": "2026-10-17T09:48:55.164Z"
}

loadTimeMs is navigation time. totalTimeMs includes the settle wait, cookie handling, scrolling and encoding. For PDFs width and height are null.

A failed URL looks like this (this one is a real 404; the image of the error page is kept for debugging but is not charged):

{
  "url": "https://github.com/this-page-does-not-exist-davi-test",
  "ok": false,
  "httpStatus": 404,
  "error": "HTTP 404",
  "screenshotUrl": "https://api.apify.com/v2/key-value-stores/<storeId>/records/008-github-com-this-page-does-not-exist-davi-test.png"
}

Input reference

Field	Type	Default	Notes
urls	array	["https://apify.com"]	Required. Up to 1,000. A URL without a scheme gets https://. Only http and https; localhost and private ranges are rejected.
device	enum	desktop-1366	desktop-1366, desktop-1920, laptop-1440, iphone-15 (393x852 at 3x), pixel-8 (412x915 at 2.6x), ipad (820x1180 at 2x). Mobile presets send a mobile user agent and touch.
viewportWidth, viewportHeight	integer	from device	Override the preset. Width 200 to 3840.
fullPage	boolean	true	false captures only the first viewport.
format	enum	png	png, jpeg, webp, pdf. PDF uses the print layout (A4), not a screenshot of the screen.
quality	integer	80	JPEG and WebP only.
selector	string	none	Capture only the first match. No match: the URL fails, not charged.
clip	object	none	{"x":0,"y":0,"width":1200,"height":630} in CSS pixels. Ignored if selector is set.
waitUntil	enum	load	load (plus 1.5 s settle), domcontentloaded, networkidle.
waitForSelector	string	none	Must become visible before capture, within the page timeout.
delayMs	integer	0	Extra wait before capture, max 30,000.
scrollToLoad	boolean	true	Lazy-load scroll before a full-page capture.
removeCookieBanners	boolean	true	See above.
hideSelectors	array	none	CSS selectors set to display:none.
blockAds	boolean	false	Aborts requests to common ad and tracker domains.
fixStickyHeaders	boolean	true	Full page only.
darkMode	boolean	false	Sets prefers-color-scheme: dark. Only sites that support it change.
pageTimeoutSecs	integer	45	Per URL, 10 to 180.
maxConcurrency	integer	3	Pages in parallel.
returnBase64	boolean	false	Adds imageBase64 to the row when the file is 2 MB or less. For agents that cannot fetch a URL.
proxyConfiguration	object	off	Apify Proxy. Choosing the RESIDENTIAL group adds the residential-page charge.

Calling it from code

JavaScript:

import { ApifyClient } from 'apify-client';

const client = new ApifyClient({ token: process.env.APIFY_TOKEN });
const run = await client.actor('pxdata/website-screenshot-api').call({
    urls: ['https://example.com', 'https://apify.com'],
    format: 'webp',
    quality: 75,
});
const { items } = await client.dataset(run.defaultDatasetId).listItems();
for (const item of items) console.log(item.ok, item.url, item.screenshotUrl);

Python:

import os
from apify_client import ApifyClient

client = ApifyClient(os.environ["APIFY_TOKEN"])
run = client.actor("pxdata/website-screenshot-api").call(
    run_input={"urls": ["https://example.com"], "device": "iphone-15", "fullPage": True}
)
for item in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(item["ok"], item["url"], item["screenshotUrl"])

cURL, synchronous, returns the dataset rows:

curl -X POST "https://api.apify.com/v2/acts/pxdata~website-screenshot-api/run-sync-get-dataset-items?token=$APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"urls":["https://example.com"],"format":"jpeg"}'

Files in the key-value store are stored under kvKey and also reachable from screenshotUrl. Unnamed storages are removed after the retention period of your plan, so copy anything you need to keep.

Pricing

Pay per event. You pay for results, not for compute.

Event	Charged when	FREE	BRONZE	SILVER	GOLD
capture	One URL returned ok: true	$0.004	$0.003	$0.0025	$0.002
residential-page	Added to each successful capture when the RESIDENTIAL proxy group is selected	$0.02	$0.02	$0.02	$0.02

Apify also charges its standard apify-actor-start event once per run. At the time of writing the platform docs list it at $0.00005 per start.

Not charged: HTTP 4xx and 5xx responses, timeouts, DNS and TLS errors, selector not found, invalid URLs, URLs skipped because the run hit its maximum cost.

1,000 screenshots on the BRONZE tier cost $3.00. Set "Maximum cost per run" in the Console if you want a hard cap; the Actor stops capturing when it is reached.

Memory and performance

Measured on a batch of 10 URLs (BBC News, The Guardian, Le Monde, Mercado Livre, Wikipedia's Brazil article, apify.com, github.com via an http redirect, globo.com, example.com and one 404), full page PNG, desktop preset, 3 pages in parallel:

• 9 of 10 captured. The tenth was the 404 and was reported as such.
• Mean 11.7 s per URL from navigation start to stored file, 57 s for the whole batch (about 5.7 s per URL with 3 in parallel).
• Fastest: example.com, 2.3 s. Slowest: Wikipedia and globo.com, 26 s and 27 s, because of their length and lazy loading.

That run was on a home connection, not on Apify's datacenter, so treat the timing as an order of magnitude.

Memory: use 2 GB (the default). A cloud test with 1 GB, maxConcurrency 2 and the same 10 heavy URLs as full page PNG did not finish: after 15 minutes only 2 of 10 pages were captured and the run hit the timeout, because Chromium was starved and concurrency stayed at 1. 1 GB is only reasonable for light pages, one at a time. Use 2 GB for up to 4 parallel pages and 4 GB for up to 8. Long pages in PNG at mobile pixel ratios are the memory-heavy case; switch to JPEG or WebP, or lower maxConcurrency, if a run is killed for memory.

Limits

• Height: 16,383 device pixels. With iphone-15 (3x) that is about 5,460 CSS pixels of page.
• 1,000 URLs per run, 180 s per URL at most, 30 s extra delay at most.
• No cookies, no logins, no custom headers. This keeps the Actor simple to use from agents and it is not built for pages behind authentication.
• Sites that block datacenter traffic return 401, 403 or a challenge page. Those rows come back with ok: false and are not charged. The Actor does not try to defeat bot protection. For sites that allow it, the RESIDENTIAL proxy group may help.
• Consent banners that live inside cross-origin iframes may survive. Open a ticket with the URL.
• Animated content is captured at one instant. Videos show their poster or first frame.

FAQ

Why is my image shorter than the page? The 16,383 px cap. Check truncated in the row. For very long pages capture in pieces with clip.

Why a 401 or 403 on a site that works in my browser? The site is blocking automated traffic or datacenter IPs. The row has ok: false, you are not charged, and the stored image shows what the server returned.

Does it log in, read paywalled content or bypass anything? No. It loads the public page as an anonymous visitor.

Can I get WebP? Yes. format: "webp". Chromium does not produce WebP in headless mode here, so the Actor captures PNG and encodes it with sharp.

The cookie banner is still there. Send the URL and the banner's selector. Meanwhile add the selector to hideSelectors.

Is it legal? Taking a screenshot of a public page is technical, not a legal green light. Copyright, terms of service and privacy law (GDPR, LGPD) apply to what you do with the images, especially if they show personal data. You are responsible for your use.

Which URLs are rejected? localhost, .local, .internal and private IP ranges, and any scheme other than http and https.

What happens when I hit the maximum cost per run? Captures already done stay in the dataset. Remaining URLs are written as ok: false with an explanatory error and no charge.

Changelog

1.0.0 First release. PNG, JPEG, WebP and PDF; six device presets; full page with lazy-load scroll; element and rectangle crop; cookie banner removal; sticky element handling; dark mode; ad blocking; optional Apify Proxy; per-URL error rows; pay per successful capture.

Support

Open an issue on the Actor page with the URL, your input and the run link. Reproducible cases are looked at first.