Translate the text in an image, keep the design
The lines are read and translated for you. Check the wording, and the image comes back with the new language in the same place and style.
Edit the text in your image
Before and after
Before
After
Before
After
Before
After
Before
AfterIn short
Google Lens and camera translators show a translation on top of the image; they do not give you a usable file. Make It Say does: pick a language and every line is translated and filled in for you. You read and correct the wording first, then the original lines are removed and the translations are drawn in their place, in the matching style.
How it works
- 1
Upload the image
PNG, JPEG or WebP. The text in it is found and boxed in a few seconds. No account needed for this step.
- 2
Type what each line should say
Click a box, type the new text, or translate every line into another language in one click. Fix a misread line or draw a box the scan missed. Change up to 20 lines at once.
- 3
Download the result
In 20 to 40 seconds you get a PNG with the new text in the original font, color and angle. One credit per image; your first is free.
Scripts it handles
Latin, Chinese (simplified and traditional), Japanese (kanji and kana), Korean, Arabic (right-to-left, connected letters), Cyrillic, Thai and others. Brush, calligraphic and carved styles are kept; see the Japanese festival poster and the Arabic bakery sign above.
Right-to-left lines are recognised in reading order. If the scan shows the words backwards, correct the original text in the box before running the edit.
Fitting a longer translation
English is usually longer than Chinese or Japanese. The new line is fitted into the same area, so very long translations may come out tighter. When space is short, prefer a shorter wording, or change the line in two steps.
What it will not edit
ID cards, passports, driving licences, certificates, transcripts, receipts, invoices and bank documents. The scan recognises these and the edit is refused before anything is charged. Editing them is forgery in most places, and we do not want that traffic.
Everything else that is yours to change is fine: your screenshots, your designs, your photos, images you have the rights to.
Questions
Does it translate automatically?
Yes. After the scan, pick a language and press Translate: a text model translates every line and fills in the boxes. Nothing is drawn yet. You check each line, fix names, prices or wording, or type your own translation, and only then make the image. Translating is free and does not need an account.
Can it translate a whole menu or a document scan?
Up to 20 lines per edit, which covers most screenshots, menus and posters. Documents such as certificates and receipts are refused. A long document is not what it is for.
Do I need to know the font?
No. The font, weight, color, outline and slant are copied from the pixels around the line. You only type the new words.
Does it keep the resolution?
Yes. The result is a PNG the same size as the image you uploaded, and apart from the lines you changed, the pixels are the ones you uploaded. Nothing is downscaled to fit a watermark or a free tier.
Is there a watermark?
No, on free and paid edits alike.
Does it work on a phone?
Yes. It runs in the browser on iPhone and Android. Upload from your camera roll, tap a line, type, download.
What does it cost?
One credit per image, however many lines you change. New accounts get 1 free credit. Then 2 edits for $1, 40 for $10, or 120 a month for $20. Failed edits are not charged. Pricing.
What happens to my image?
It is sent to the model once, stored for up to a day so the edit can finish, then deleted. Results are kept for 30 days so you can download them again. See the privacy policy.