Journal Submission File Size Limit: Compress Figures on Mac

It depends on the journal and the file. Publishers usually cap each figure, sometimes the whole submission, and often leave the exact number to the portal’s upload screen. These are the figures each publisher and repository states on its own page.

All rows were read 11 October 2026. Read your journal’s own instructions for authors, and the error text its portal gives when a file is rejected.

Journals state resolution in dpi at print size. Elsevier asks for 300 dpi for halftone images, 500 for combination art and 1000 for line art. PLOS ONE accepts 300 to 600 dpi, 789 to 2250 pixels wide at 300 dpi, as TIFF or EPS.

Dpi only means something against a print width. At 300 dpi, Elsevier’s 90 mm single column is 1063 pixels wide and its 190 mm full width is 2244 pixels. A 6000-pixel camera frame has more pixels than a 190 mm halftone figure needs, so resizing a review copy to the printed width is reasonable.

For the PDF side of print quality, compressing a PDF without losing 300 PPI covers how Smol’s presets behave. Presets are long-edge pixel caps, not dpi targets.

The arXiv help pages publish no numeric limit. They say oversized submissions usually come from “a small number of inefficient figures”, and that since February 2026 arXiv warns about images larger than 34 megapixels. Keep every image under that.

The arithmetic is simple: 6000 × 5600 pixels is 33.6 megapixels, while 6000 × 5800 is 34.8. arXiv’s own advice is JPEG for photographs, PDF, PNG, GIF or PostScript for diagrams, and larger versions as ancillary files. It does not convert figures for you and refuses submissions with omitted figures. PDFLaTeX needs .pdf, .jpg or .png figures, while plain LaTeX needs .ps or .eps, so match your figures to your build.

Smol helps with the raster part: set a maximum dimension, compress photographs to JPG and write the result as a copy beside the original. It does not rewrite LaTeX, so keep an eye on the portal’s warnings.

Each publisher sets its own supplementary file size limit, and none of the publisher pages cited here states one for supplements. Check the portal. Repositories are more generous: Zenodo takes 50 GB across up to 100 files per record, so large video is better deposited there and cited by DOI.

When the portal does cap supplements, the usual culprits are video and microscopy exports. Smol compresses video to MP4, MOV, WebM or MKV with H.265 as the default and H.264 as the option. H.264 plays on more devices, and our H.265 vs H.264 guide explains the trade. Lower the output resolution when quality High is still too big, and check the output size, since Smol sets quality and pixels, not bytes. Compressing 4K video has measured examples.

Resize first, then set quality. Smol resizes images to fit inside 2000 × 2000 pixels unless you change it, so a full-width figure at 2250 pixels needs the setting raised. Quality numbers below are starting points we chose, not measured thresholds.

At default settings Smol made 24 high-quality JPEG photos 69.3% smaller in our published benchmark. Those were small 768 × 512 film photographs, so figures from your microscope or camera will differ. Metadata is stripped by default, including EXIF, GPS and IPTC, which also removes camera and location details from unpublished images; see removing EXIF data on Mac.

One trap: converting to JPG ignores the quality setting. To shrink a TIFF figure for a review PDF, use Compress with output JPG, not Convert.

Do the whole folder once, then read the largest output against the cap on the portal.

1. 1

   Read the cap on your journal’s upload screen

   Write down the per-file and total limits, plus any pixel or megapixel rule.

2. 2

   Copy the figures to a review folder

   Keep production TIFF or EPS originals elsewhere. Smol writes copies with a _compressed suffix and does not overwrite originals by default.

3. 3

   Drag the folder into Smol

   Figures, a manuscript PDF and a supplementary clip can go in together.

4. 4

   Set format, quality and maximum dimension

   Use the settings table above. Keep fit mode so nothing is cropped.

5. 5

   Sort by size and open the worst file

   If the largest file clears the cap, they all do. Zoom into axis labels and scale bars before you upload.

6. 6

   Rerun the files that miss, lower

   Drop quality in steps of 5, or reduce the maximum dimension. For one image, target-file-size mode accepts 4 to 4096 KB, aims at a band rather than a ceiling, and is slow, about 28 seconds for one 768 × 512 image.

Co-authors on Gmail hit 25 MB, and encoding adds about 37%, so the PDF-for-email guide recommends aiming near 18 MB. Our scanned PDF guide covers supplements built from scans.

Free tools come first. Preview’s export, ImageMagick and Ghostscript are fine, but each has a catch. Compare them with a batch tool by what you need.

Our compared methods and Preview write-up go deeper. The thesis PDF page and the education uploads page cover adjacent jobs.

Smol is wrong for these jobs, and we would rather say so.

• The journal wants TIFF or EPS at 300 to 1000 dpi. Send your production originals. Smol re-encodes and has no lossless mode, and it cannot write EPS.
• Your PDF is mostly vector graphics. PDF presets resample images by pixel cap, so a vector-only PDF may barely shrink.
• You need a lossless or pixel-identical optimizer for published figures. ImageOptim is the answer there, see our ImageOptim comparison.
• You work on Windows or Linux, or need compression in a CI pipeline. Smol is macOS only; the Smol API covers servers.
• You need redaction or OCR. Smol does neither.

What about data privacy for unpublished results?

Smol runs on your Mac and uploads nothing, so unpublished figures never reach a third party. That is a statement about where files go, not about your institution’s rules. Follow those.