About ChangeTextInImage

ChangeTextInImage changes the words in a picture. You draw a box around the text, type what should be there instead, and an image model redraws those words to match the original lettering and background. You see a free watermarked preview first, or use credits to get the full-size image straight away.

How an edit works

1. Your picture stays in your browser until you press Change text.
2. It is uploaded, and an automated check looks at whether it's a document or a screenshot — the tool is for designs, products and photos, and a refused image is never stored.
3. The image, your box and your new text go to Black Forest Labs in the United States, whose FLUX 3 Image Klein model redraws the words.
4. A free change comes back as a small watermarked preview plus a close-up of the changed words; a change made with credits comes back at full size with no watermark.
5. When you download, you get the model's output at the size you chose — 1K, 2K or 4K — as a JPEG or PNG, without the watermark.

Typical waits: about 10 seconds at 1K, about 30 seconds at 2K, about 4 minutes at 4K.

The model

The edit is made by FLUX 3 Image Klein from Black Forest Labs. Every image it returns carries Content Credentials — a signed, invisible record inside the file naming the model and its maker and stating that the image was made with AI. We pass that file to you unchanged, so the record stays valid, and the preview carries it too, inside Cloudflare's own signed record. Platforms that read Content Credentials, such as Meta's, may show an "AI info" label when you post the image.

What it's made for

Posters and flyers, product and e-commerce images, memes, AI-generated images with a misspelt word, slides, and signs and labels: pictures of things, where a word, a date or a price needs to change.

It is tested on Latin-alphabet text. In our tests before launch, the new words came out with every letter right and no visible seam on 18 of 20 test images at 2K, allowing up to two tries each. Because the whole picture passes through the model, check the close-up that comes with every result, and try again if something needs another go.

What it declines

An automated check looks at every image before any edit is made and refuses the ones it recognises as identity documents, receipts, invoices, payslips, bank statements, certificates, medical notes, tickets, or screenshots of payments, messages or posts. The Terms also rule out misleading edits of real people's words, watermark removal, images of anyone under 18 and sexual content of real people. This keeps the tool for the people it's built for, and away from forgery. If you see an image made here being used to deceive, email support@changetextinimage.app.

Who's behind it

ChangeTextInImage is run from Singapore by a small independent studio that builds web tools; its company details are in the Terms. Payments are handled by Stripe, and the site runs on Cloudflare.

Contact

Questions, a result that went wrong, a refund, or a misuse report: support@changetextinimage.app. We read every message.