EditTextInImage

If an event heading changes after a poster has been exported, EditTextInImage offers a way to replace that phrase without rebuilding the whole picture. You draw a box around the complete text, enter its current wording and supply the replacement. The website shows before-and-after examples for posters, video covers and product graphics. Those are the maker’s examples; ProjectRadar did not generate an edit.

This is pixel reconstruction, not a recovered text layer. The model tries to match the visible style, and the browser places the edited rectangle into the original image. The site says pixels outside the selection are retained. Inside it, lettering and background can change, so correct spelling alone is not enough: edges, spacing and nearby details need inspection too. If the editable design and its font are available, they remain the more predictable route.

The editor accepts opaque, static JPG, PNG and WebP files up to 10 MiB, 6,000 pixels on either edge and 12 megapixels. Transparent and animated images are not supported. The intended task is one short, clear phrase, not a page of text or a layout redesign. The download keeps the original dimensions as a PNG, but it cannot restore detail missing from the source.

A verified account receives two free credits once, with no daily reset. Purchases are closed, so editing stops when those credits run out. The terms say a confirmed model failure returns a credit; an unsatisfactory result or cancellation does not automatically do so.

The privacy policy says the full original stays in the browser, while the selected crop, nearby context and supplied wording go through the service to Replicate. Provider-hosted crops and outputs follow Replicate’s retention rules. Local preview removal does not guarantee immediate provider deletion or cancellation of work already running.

Results are temporary, and there is no permanent account gallery. Download a usable result before closing the tab, and retain the original for comparison.