YouTube thumbnail enhancer: sharper, higher-resolution thumbnails

Most thumbnails start as a frame grabbed from the video. Video is compressed much harder than photos, so a frame that looks fine in motion shows smeared detail, blocky shadows and soft edges when it is paused and saved. Frames from 720p or older footage are also small, so YouTube has to enlarge them on big screens.

Other common sources are old thumbnails you want to reuse at a higher resolution, photos a guest sent through a chat app, and screenshots of a camera preview. Each of these has been reduced or compressed somewhere along the way.

YouTube's help centre lists the custom thumbnail requirements shown in the table: 3840 x 2160 pixels for videos with a minimum width of 640 pixels, a 16:9 aspect ratio, JPG or PNG, and a file size limit that depends on the device you upload from. Check the help page before you upload, because these figures have changed before.

In practice a 1280 x 720 frame is well above the minimum but below the recommended size. A 4× enlargement takes it to 5120 x 2880, which you can reduce to 3840 x 2160 in any editor.

For a 720p or 1080p frame, use Upscale 4×. Both sizes are within its 5 MP input limit: a 1280 x 720 frame becomes 5120 x 2880 and a 1920 x 1080 frame becomes 7680 x 4320, which you then reduce to 3840 x 2160. It is built for photos of people, places and objects and adds natural fine texture so the enlargement does not look smeared.

For a very small source, such as an old 480 x 360 thumbnail, Upscale 8× reaches 3840 x 2880 in one run. It accepts images up to 1.25 MP and uses 2 credits per photo.

If the frame is grainy because it was filmed in a dim room, or soft and blocky from video compression, run Enhance on it first. It removes noise and compression damage at the same size, and you can then enlarge the result.

Pick the sharpest frame. Step through the video a few frames at a time around the moment you want; frames in the middle of fast motion are blurrier than the ones just before or after. Export the frame from your editor at full resolution instead of screenshotting the player.

Run the frame through the model that fits, then open the result in your design tool. Add text, arrows and graphics after enlarging, not before. Text drawn at the final size will always be sharper than text that was drawn small and enlarged. Export as JPG or PNG within YouTube's size limit for your device.

Faces are the most important part of many thumbnails, and they are where generative tools cause trouble: eyes change shape, skin turns waxy, and the result no longer looks quite like the person. PhotoRescue is not a face generator and does not swap in a new face. Still, check eyes and mouths at 100 % before you publish.

Text already in the frame is kept, not rewritten. Small text stays as soft as it was, because the models do not guess letters. If a word matters, type it again as an overlay in your design tool.

Frames with strong motion blur, a subject badly out of focus, or a face that is only a few pixels tall will improve only a little. For those, choose a different frame, take a still photo for the thumbnail, or build the design around graphics.

Free results carry a watermark, which you will not want on a public thumbnail, so test a few frames on the free plan. Pro covers 200 credits a month for $9.99 and Pro+ 500 for $19.99, both without watermark and with batches of up to 50 images, which helps when you refresh a back catalogue.