How to Place Text Behind a Subject in an Image

It is a poster tool. It is not a compressor, not a converter, and not a studio. If the photo does not have a subject, the effect has nothing to hide the letters under.

The photo has to do half the work

A talking-head frame, a product held in one hand, a dog in the middle of the shot: the silhouette is obvious. The cutout library can grab it. A conference group shot, a landscape, or a white-background pack of cables: there is no “behind.” The model will guess, and the guess will be a chewed edge or a missing ear.

Hair and glasses are the usual failures. A backlight halo around a head will either eat the hair or leave a rim of the original background in the type. If the first mask is wrong, change the photo. Fighting a bad cutout for an hour is how you miss the upload window.

Use a high-enough resolution that the subject is sharp at the crop you will actually ship. A 640-pixel selfie upscaled to a 1280-wide thumbnail looks like a 640-pixel selfie with fancy type.

Type is a layout problem, not a font buffet

Pick one short line. “NEW VIDEO” or the episode title, not a paragraph. Heavy weight, high contrast against both the background and the subject. White type on a white shirt behind a pale wall is invisible even if the layering is perfect.

The first run of the tool may fetch a cutout model into the tab. That is a library download, not your photo leaving to be stored. After that, you set the words, size, and position. The letters belong mostly in the empty part of the frame so the subject covers only the edges of a few glyphs. If every letter is buried, nobody can read the thumbnail at 120 pixels tall in a sidebar.

YouTube will crop. Instagram will crop harder. A podcast square will crop off the sides of a 16:9 export. Look at the download inside the shape the host uses. If the face is gone, you designed for the wrong rectangle.

A thumbnail you can finish in one sitting

You have a still of a person holding a broken hard drive. You want the word RECOVERED to sit in the wall behind them, with the drive and the hands in front. You drop the still on Text Behind Image, wait for the cutout, set a condensed heavy face, push the type large, drop it so the last letters duck under the sleeve. Export. Open the PNG at 1280×720. If the crop still shows the face and the word, you ship it. If the file is chunky for the upload, run that PNG through the compressor once - do not JPEG the type into mush.

The still never needed to sit in a “free thumbnail maker” account that keeps a gallery of other people’s faces.

Font choice matters less than contrast and size. A condensed black grotesque at 200 pixels will read in a sidebar. A script face at 40 pixels will not, even if the layering is clever. Load one heavy family, set the line, and stop shopping the font list.

When to stop

The mask took the lamp and left the person. Stop. New photo.

You need brand-perfect kerning and a grid. Use a real layout tool. Bring a PNG cutout if you want; this page already gave you the idea.

You need every frame of a video captioned. That is a caption workflow, not a still effect.

The image is a client’s unreleased product or a kid. Local processing is the point. Do not also drop the export into a random “enhance thumbnail” site.

Color pairing with a UI is a different job - the palette article if you are matching a site, not a face. For the rest of the local image tools, use the roundup.

When you have a still with a real subject and a few words that should sit in the room, open Text Behind Image, check the crop the platform will apply, compress if you must, and leave.