We have all been there: you have a perfect product shot or a crucial screenshot, but there is an expired 'SALE' badge, a burned-in caption, or a user name in the UI chrome that needs to go. Since the text is already flattened into pixels, there is no text layer to delete. Blur bars look amateur, and manual clone-stamping is painfully slow. A new tutorial from DEV.to user jaweii outlines a streamlined workflow using AI inpainting to remove text from images online, repair the background, and optionally replace the wording.

What 'Removing Text' Actually Means

The guide clarifies a common misconception: when you search for 'remove text from image,' the letters are already part of the pixel data. A production-ready cleanup requires three steps: detecting the lettering via OCR or region detection, erasing the strokes along with attached shadows or outlines, and repairing the background so that textures like sky, fabric, or UI panels continue naturally. This is AI inpainting aimed specifically at text, not just a simple 'delete' button. Flat walls and gradients rebuild cleanly, but faces and fine patterns may require a second pass.

The Tool: PhoText

The tutorial recommends PhoText, a freemium, browser-based AI editor designed for text baked into images. The process involves uploading a JPG, PNG, or WebP file, selecting the words, and letting the model rebuild the fill without any software installation. Beyond removal, it offers click-to-edit replacement, font recognition, and watermark removal for assets you own. The author notes that while free tiers exist, specific pricing and limits change frequently, so users should check the live product for current plans.

Step-by-Step Workflow

1. Upload the sharpest export: Use the original file rather than a photo of a screen to ensure better detection. 2. Select the text: Click the badge or label to remove it, using manual tools if AI detection misses a glyph. 3. Erase and repair: Run the erase path to clear strokes and rebuild the background from neighboring textures. 4. Optional replacement: Type new text in the same spot and use font identification to match the style. 5. Use conversational AI for complex backgrounds: For textured items like wood grain, describe the edit in plain English (e.g., 'remove the SALE badge and keep the fabric weave'). 6. Inspect and export: Zoom to 100% to check for artifacts, then download as PNG for lossless masters or high-quality JPEG for the web.

Key Takeaways

  • Check your permissions: Only edit images you created, own, or have written permission to change. Do not use this to forge IDs, certificates, or strip watermarks from stock assets you don't control.
  • Know the limits: AI rebuilds from surrounding pixels; it cannot recover what was never captured. It is best for marketing and creative assets, not for print-production accuracy or legal documents.
  • Privacy matters: When removing names or emails from screenshots, follow your team's policy on sensitive data. Zoom in to ensure nothing readable remains.

The Bottom Line

This workflow is a game-changer for developers and support writers who need to clean up UI screenshots quickly. While it won't replace professional graphic design for complex composites, it handles the cases where you just need to get rid of a distracting badge or redact sensitive info without opening a heavy desktop app.