How to Make a YouTube Thumbnail That Reads at Small Sizes

1280×720 under 2 MB is the easy part. The hard part is the small size: 150 px text ends up 26.2 px tall in a 320-wide result, and white text over a photo went from 3.87:1 to 5.65:1 contrast with a 20% darken.

The Specs That Matter

YouTube’s help page on custom thumbnails recommends 1280×720 pixels with a minimum width of 640, and a file under 2 MB for video thumbnails (10 MB for podcasts). Two details catch people out. First, the 2 MB limit is about the file, not the pixels: a 1280×720 PNG of a photo can be several times the size of the same image as a JPG. In our tests the same photo design came to 1,005,395 bytes as PNG and 210,561 as JPG, and a background of pure random noise, about the worst case there is, made a 2,103,985-byte PNG (over the limit) but a 679,744-byte JPG. Second, for vertical videos YouTube replaces a 16:9 custom thumbnail with an automatic 4:5 one on the home, explore and subscription pages, so a landscape thumbnail is only fully in your control for landscape videos.

Design for the Small Size

A thumbnail is designed at 1280 pixels wide and mostly seen far smaller. Scale is unforgiving: in our tool’s default font, 150 px text has capital letters 105 px tall at full size, about 26.2 px tall at 320 pixels wide and 13.8 px at 168 wide. Anything that starts as body-size text in the design is gone by then. That is why the working rules are blunt: three to five words, the heaviest weight the font has, one face or object large enough to read on its own, and nothing important in the bottom-right corner, where YouTube shows the video’s length.

The quickest check is to look at your design at the small size before you export. Our YouTube Thumbnail Maker draws two small previews with a sample time stamp next to the editor for exactly that reason.

Contrast Against a Photo, Measured

White text over a photo is the classic thumbnail look, and the photo decides whether it works. We measured the WCAG contrast ratio of plain white text (no outline) against the average color of the photo pixels behind it, with the tool’s Darken setting at four levels:

Darkening the photo by 20% took the text from below the 4.5:1 level used for normal text in WCAG to comfortably above it. A thick dark outline does a similar job without touching the photo, because every letter then sits on its own dark edge; that is why the outline is the default in our tool and why its contrast is shown separately.

Cut-Out Subjects

A face or product cut out from its background and placed over a simple backdrop reads better at small size than the same subject inside a busy frame. Cut it out first (our Background Remover runs in the browser and saves a transparent PNG), then add a solid outline, usually white, so the edge of the subject stays visible over any background.

Exporting Without Surprises

Export at exactly 1280×720 and as JPG unless you need PNG for a reason. Our tool exports JPG at quality 95 and, if a file would be over 2,000,000 bytes, steps the quality down until it fits; we read the limit as decimal megabytes so the file passes whichever way YouTube counts. In Chromium no design we tried needed the step-down, because the browser’s JPEG encoder at quality 95 kept even the random-noise test at 679,744 bytes.

Frequently Asked Questions

Published Oct 14, 2026 by Derek Giordano