Image Size for News Website Photos, and How to Compress Press Photos

For an article lead, 1,200 to 2,000 pixels wide, as a JPEG or WebP, is the working range. Google asks for 1,200 pixels as a minimum for large Discover images, and anything beyond what the template displays only adds weight.

The sourced figures are pixel sizes and load time, so set a kilobyte budget for your own pages. The limits are in the table below, along with the page-speed target: a good Largest Contentful Paint is 2.5 seconds or less at the 75th percentile, and poor is above 4.0 seconds. When the lead photo is the LCP element, its bytes are the first thing to cut. Our Core Web Vitals image guide covers the metric in more depth.

Google, WordPress and web performance guidance publish the figures below. Your own CMS may add stricter ones, so open your upload screen and read its error text.

Self-hosted sites have their own server limits, so there is no single number. If a reporter’s upload fails, read the error text, since a smaller file is often the quickest fix.

Aim for the smallest file that still looks right at the largest size your template shows it, and check the result. Treat the settings below as starting points, not thresholds.

Pixels matter more than quality. A 6000-pixel frame fitted inside 2000 pixels keeps about a ninth of its pixels, which is why resizing beats lowering quality for large camera files. In our benchmark, Smol made 24 high-quality JPEG film photos at 768 × 512 69.3% smaller at default settings. Those photos are small, so a camera file will not follow the same percentage.

Use Compress, not Convert, so quality applies. Set the output format and a maximum dimension, keep the resize mode on fit so nothing is cropped, and write the numbers on the desk’s checklist.

AVIF is worth a test for leads: with its lossy mode on and matched for quality, Smol’s AVIF was 20 to 22% smaller than ImageOptim’s JPEG on 24 small test photos (see the data). Check that your CMS and its CDN serve it, then keep a JPEG fallback.

Do you need several sizes of one photo?

One Smol run produces one size. Responsive sets with srcset are usually made by the CMS or an image CDN from your single upload, which is why a clean 2000-pixel original is a better upload than a heavily compressed one. Making three widths by hand means three runs and renaming, since the output name is the same each time.

What about credit lines and captions?

Metadata stripping is on by default and removes EXIF, GPS, IPTC and XMP, which includes copyright and contact fields. If the credit and caption must stay in the file, turn it off for those jobs. For published web copies, stripping removes camera serials and GPS; read more in does compressing remove metadata, and see our page for photojournalists and metadata.

A photo desk workflow with Smol is one drop per batch: copy the selects, compress them to the desk’s preset, check the largest file, then upload to the CMS. It takes the same steps for ten photos or a hundred.

1. 1

   Copy the selects out of the ingest folder

   Keep the full-size originals where they are. By default Smol writes a copy with a _compressed suffix and does not overwrite the original.

2. 2

   Drag the folder into Smol

   Whole folders work, and mixed batches are fine: photos, a PDF release and a clip can go in together.

3. 3

   Set JPG, quality 75, maximum dimension 2000

   Raise it to 2560 for feature leads. Turn metadata stripping off if credits must travel in the file.

4. 4

   Compress, then sort the output by size

   Open the largest file. If it is heavier than your page budget, lower quality to 70 or the dimension to 1600 and rerun only that file.

5. 5

   Upload to the CMS and read the page

   Check the live article on a phone connection. The CMS may scale or re-encode again, so judge the published page, not the file.

Save the settings as a preset. Presets live on one Mac, so a desk with several machines writes the numbers down and recreates the preset on each. A watched folder can run the job for you, but only point it at folders of fresh, unoptimized exports, because it can make already-optimized files larger; details in the watched folder guide.

Licensing for a desk: Solo is $29 once for one Mac, Team is $49 once for three Macs, and Enterprise is $20 per seat, once, for 5 to 500 seats. There is no subscription.

Resize first, then lower quality in small steps. Press photos hold faces, text on signs and fine texture, so look at the 100% view of a crowd shot or a document close-up before approving a setting.

• Quality 75 up to 80 for photos with faces and skin, where banding shows first.
• Keep screenshots, charts and graphics as PNG or compress them as PNG rather than JPG, where text edges smear.
• Smol has no lossless mode for images: it always re-encodes. Archive masters stay untouched, and only web copies are compressed.
• Camera RAW (CR2, NEF, ARW, DNG) is accepted as input, but Smol does not develop it. Export a JPEG first.

For a wider look at the options, read image optimization for web developers, or our compress images for web page. If your team runs a WordPress site, the WordPress image page covers the CMS side, and image SEO and page speed goes further on ranking.

Free options exist, and they work for one-off jobs. Smol earns its place when the desk repeats the same batch every day and wants one setting applied to a whole folder.

For comparisons, see the sips command guide, Squoosh alternatives and Smol vs TinyPNG. The last matters for embargoed material: an online compressor receives your file on its server, and Smol does not upload.

Smol is a desktop app for people at a Mac. It is not the answer in these cases.

• Automatic processing on every CMS upload. Smol is not a build-time or server tool. For a pipeline, use your CMS’s image handling, an image library, or the Smol API.
• Pixel-identical optimization. Smol always re-encodes. For lossless savings, see the ImageOptim head-to-head.
• Exact aspect-ratio crops. Fill mode crops with no control over what is cut. Crop in your photo editor, then compress.
• Windows or Linux desks. Smol is macOS only.
• SVG graphics, JPEG XL or developing RAW. None are supported.
• Redaction. Smol does not redact documents or blur faces.