Last reviewed: November 9, 2026

remove text from image tool

Neither free tool removed the large opaque lettering in this test: unwatermark.ai erased just 1 of 3 text blocks and returned a 434,830-byte 1224 x 816 JPEG, while phototune.ai erased none of the three.

One 2400 x 1600 JPEG (3.84 MP, 1,613,250 bytes) with three opaque text blocks — a top-left dark banner, a center wordmark and a bottom-right credit line — sent unchanged to both tools on November 9, 2026.
Tool	Text blocks erased (of 3)	Output size	Download bytes	Format	Sign-up / card
unwatermark.ai	1 of 3 (bottom credit line only)	1224 x 816 px (1.00 MP)	434,830	JPEG	Neither required
phototune.ai	0 of 3	2400 x 1600 px (3.84 MP)	4,186,735	PNG	Neither required

If the target is a small corner credit line and a 1224 x 816 JPEG at 1.00 MP is acceptable, pick unwatermark.ai, the only one of the two that erased any text; for a large opaque banner or center lettering neither download is usable, so use a manual brush editor such as remove text from image instead.

What each column means for you

Tool. This column names the free service behind the row, so every other cell can be matched to the site you would actually open.
Text blocks erased (of 3). This is the only column that answers whether the lettering you wanted gone is gone, so read it before any other cell.
Output size. This tells you how many pixels survive the free download, which decides whether the file still holds up for print or a tight crop.
Download bytes. This is the weight you will upload, store or email, and it separates a light JPEG from a heavy PNG.
Format. This tells you whether the download is a lossy JPEG or a lossless PNG, which matters if you plan to edit the result again.
Sign-up / card. This tells you what has to be handed over before the file arrives, so you know whether the job stays anonymous.

Measured result for remove text from image tool: unwatermark.ai erased 1 of 3 opaque text blocks and returned a 434,830-byte JPEG, phototune.ai erased 0 of 3 and returned a 4,186,735-byte PNG. — One 2400 x 1600 JPEG with three opaque text blocks, pushed through both free tools on November 9, 2026.

Which tool to pick, plus the three text regions with measured black-ink coverage and whether each block was removed. — The recommendation and the per-region check, using a fixed rule: more than a 60% drop in near-black ink means removed.

How the two downloads were checked

The same single input image was pushed through each tool, and the real downloaded files were measured, then resampled to 2400 x 1600 and compared crop-by-crop against the input.

What opaque text does to automatic detection

Automatic removal works best on lettering that is thin, light, or blended into the background, because the repair can rebuild the covered pixels from the detail around them. Opaque text is the opposite case. A solid dark banner, a heavy center wordmark and a credit line burned into a corner replace the pixels underneath completely, and those pixels are not stored anywhere in the JPEG. A detector can still find the lettering, but the repair has no original color to restore, only an estimate built from the edges of the block. That is why a download can look processed and still carry the banner and wordmark unchanged, and why the small credit line was the block that actually came out in this test.