Photo translator / Private image translator
What goes out, and what does not
A private image translator: your file stays on your machine.
The reason people hesitate before dropping an image into a translation tool is never curiosity about the output. It is an unreleased price, a product that has not launched, a screenshot with a customer's name in it, or a contract someone photographed. This page says precisely what this tool sends over the network, what it does not, and how to check both without taking anyone's word for it.
The one request
When you press Detect text, a downscaled JPEG copy of that image is sent to an OCR service to find where the text sits, and it is not stored. That is the whole list. It happens once per image, only after you press the button, and never on upload.
What does not go out:
- The full-resolution original. It is decoded into a canvas in your tab and stays there.
- The finished exports. The PNG per language, and the ZIP of all of them, are assembled in the browser and handed to your download folder directly.
- Your text — original or translated. This site calls no translation API and no AI service at all. Not one character.
- Filenames, folder names, or anything else about the file.
Why the translating is your job, not ours
Most tools in this category translate for you, which means your strings go to their server and on to a model provider. This one hands you a prompt instead: it copies a table already filled with the text found in your image and opens your own ChatGPT tab. You paste, it answers, you paste the answer back. The round trip is between you and a service you already chose; this site is not in the middle of it and has no key of its own.
If you would rather not involve a model at all, download the CSV, Excel or JSON template, fill it with copy your team already approved, and upload it back. That path touches nothing but your browser.
Check it yourself in about a minute
None of this needs to be taken on trust — it is observable from the browser you are reading this in. Open DevTools, select the Network tab, then:
- Drop an image in. No request appears. The file was never uploaded.
- Press “Detect text”. Exactly one request appears, to the OCR endpoint. Its payload is the downscaled copy.
- Press “Translate with ChatGPT”. No request appears — the prompt went to your clipboard and a new tab opened. Paste the reply back: still nothing.
- Export. No request appears. The PNGs are built and saved locally.
What you will also see, and what is not about your file: the page's own assets, and two anonymous page-view pings — one to Google Analytics, one to Cloudflare's beacon. Both are sent on every page of this site and know nothing except which URL you are on.
What "runs in your browser" actually buys you
It is not a policy promise, which is the point. A retention policy is a statement about what a company intends to do with a file it already has. Rendering in the browser means the file is not there to have. There is no upload queue to be breached, no bucket to be misconfigured, no retention window to expire, and no account tying the work back to you — there is no sign-up at all.
The honest limit: text detection needs the network, and a downscaled copy of the image is what it needs. If even that is too much for a given file, draw the boxes by hand on the preview and type the original text in — then the entire run happens on your machine, and the tool never makes a request at all.
Questions
Is it safe to upload images to an image translator?
It depends entirely on where the work happens. If the tool renders the translated image on its servers, your file is uploaded, queued and usually retained for some period — read the retention clause before you send anything unreleased. Here the rendering happens in your browser, so the only thing that goes out is a downscaled copy used to locate the text.
What exactly leaves my machine?
One request, when you press “Detect text”: a downscaled JPEG copy of that image, sent to the OCR service to find where the text sits. It is not stored. The full-resolution original never goes out, the finished exports never go out, and no text — original or translated — is ever sent anywhere by this site.
How can I verify that myself?
Open DevTools, go to the Network tab and use the tool. Loading an image produces no request at all. Pressing “Detect text” produces exactly one, to the OCR endpoint. Translating produces none — the prompt goes to your clipboard, and you paste it into your own ChatGPT tab. Exporting produces none.
Can I use it offline?
Everything except text detection works offline: decoding, drawing boxes, typing translations, re-typesetting and exporting all run locally. OCR is the one step that needs the network. If you draw the boxes and type the text yourself, the whole run happens on your machine.
Is there an account, and is anything stored?
There is no sign-up and no server-side storage of your work. The boxes found in an image are cached in your own browser's localStorage so a reload does not lose them; clearing site data removes them. The site does count anonymous page views — Google Analytics and Cloudflare's beacon — and both see the URL and nothing about your file.