How to Translate Text from a Photo (The Traveler's Secret Weapon)

---

The first time I stood in front of a restaurant menu in Tokyo with almost no English on it, I genuinely considered just pointing at random items and hoping for the best. The waiter was patient, the menu was four pages long, and I had no idea what anything was. A friend next to me pulled out his phone, pointed it at the menu, and watched an English translation appear on top of the Japanese text in real time. I felt like I’d been living in the Stone Age.

That was years ago, and the tools have gotten a lot better since then. But there’s still a gap between “this works perfectly in a demo” and “this actually works when I’m standing on a street corner in Istanbul with no data.” Here’s what I’ve learned about translating text from photos while traveling—what works, what doesn’t, and how to set it up before you leave so you’re not stuck squinting at a pharmacy label at midnight.

The menu problem (and why it’s bigger than menus)

Restaurant menus are the obvious case, but this comes up everywhere when you’re abroad:

• Street signs and directions. You’re trying to find an address and every sign is in a script you can’t read. Google Maps helps, but sometimes you need to read the actual sign to figure out which building entrance to use.
• Medication and product labels. You have a headache and you’re standing in a pharmacy looking at a box of pills in Korean. You need to know the dosage and active ingredients before you take anything. This is one where a bad translation can actually matter.
• Transit tickets and machines. Train ticket machines in some countries have a language option, but many don’t—or the English translation is worse than running OCR on the original and translating yourself.
• Official documents and forms. Rental agreements, bank forms, visa paperwork. You want to know what you’re signing.
• Supermarket products. Is this soy milk or regular milk? Is this shampoo or conditioner? When you can’t read the label, shopping takes three times longer.

The flow is always the same: get the text out of the image (OCR), then translate it. The tools differ in how they chain those two steps together—and whether they work without an internet connection.

Three ways to translate text from a photo

1. Google Translate camera — real-time overlay

Open Google Translate, tap the camera icon, and point your phone at the text. The app detects the text, translates it, and overlays the translation right on top of the original—like augmented reality for language. It’s genuinely impressive the first time you see it work on a busy menu or a street sign.

When it’s great: Menus, signs, labels—anything with clear, printed text in a major language. The real-time overlay means you don’t even have to take a photo. Just point and read.

The catch: It needs an internet connection for the live overlay to work well. On a spotty connection, the overlay flickers and the translations lag or drop out. Without any connection, it won’t work at all—unless you’ve downloaded the language pair beforehand.

The fix: Before your trip, open Google Translate, go to Settings, and download the language pair you need (like Japanese → English or Turkish → English). The download is usually 30–50 MB per language. With the language downloaded, you can still use the photo mode offline—take a picture, and the app will translate the text in the photo without data. You lose the live overlay, but you get a workable translation. I always download the languages before boarding the plane. Wi-Fi at foreign airports is unreliable, and by the time you need the translation, you need it right then.

2. Textora + any translator — scan on-device, translate anywhere

This is the two-step approach. First, use Textora to scan the photo and extract the text on your device—no internet required. Then copy the extracted text and paste it into whatever translator you prefer: Google Translate, DeepL, Apple Translate, or anything else.

Why bother with two steps? A few reasons. The OCR step runs entirely on your phone, so you’re not sending images of your medical prescriptions or financial documents to a server. You also get the raw text, which means you can paste it into multiple translators and compare results—useful when one translator gives you something that doesn’t make sense. And if you don’t have data right now but will have Wi-Fi later, you can scan the text at the restaurant and translate it back at the hotel.

I used this approach a lot in rural Turkey, where I had intermittent cell service. I’d photograph menus and signs throughout the day, extract the text right away, and then batch-translate everything when I got back to the hotel Wi-Fi. Not as instant as the Google camera, but reliable.

Tip: If you’re scanning a menu, take the photo of one section at a time rather than the whole page. OCR accuracy improves when the text fills more of the frame, and you get cleaner results to paste into the translator.

3. Apple Live Text + Translate — built-in and quick

On iPhone (iOS 16+), you can open a photo, tap and hold on the text, and Live Text will let you select it. Once selected, tap “Translate” from the popup menu. If the system supports that language, you get an inline translation without installing anything.

The upside: Zero setup if you’re in the Apple ecosystem. No third-party app needed. It can work offline if you’ve downloaded the language in the Apple Translate app (Settings → Translate → Downloaded Languages).

The downside: Language support is more limited than Google Translate—Apple supports fewer language pairs, and some of the ones it does support have weaker translations. Handwritten text or decorative fonts may not trigger Live Text at all. And the “Translate” option doesn’t always appear in the popup menu for every language, which is confusing when you’re counting on it.

For major European and East Asian languages with clear printed text, it works well and it’s fast. For less common languages or messy handwriting on a chalkboard menu, you’ll probably need one of the other methods.

What works offline (and what doesn’t)

This is the part that trips people up. “Does it work offline?” is actually two separate questions:

Step 1: Getting text from the image (OCR). This can be fully offline. Live Text runs on-device. Textora runs on-device. Most decent OCR apps process locally. So you can always go from “photo” to “text” without data.

Step 2: Translating the text. This needs a connection unless you’ve downloaded the language beforehand. Google Translate has offline packs for about 60 languages. Apple Translate has fewer. DeepL has limited offline support.

So the practical setup for “no data” travel is:

1. Download your target language(s) in Google Translate before you leave
2. Have an on-device OCR app ready
3. Photograph → extract text (offline) → paste into translator (offline with downloaded language)

If you didn’t download the language and you don’t have data, you’re stuck. I learned this the hard way at a bus station in rural Romania. No data, no Wi-Fi, and I hadn’t downloaded Romanian. I could scan the schedule sign and get the text, but I couldn’t translate it. A stranger noticed me squinting at my phone and just told me which bus to take. Worked out fine, but I downloaded every language I might need after that.

Languages and scripts that trip up OCR

OCR handles some scripts much better than others. Here’s what I’ve noticed:

• Latin alphabet languages (English, Spanish, French, German, etc.) — Work well across all apps. High accuracy on printed text.
• CJK (Chinese, Japanese, Korean) — Quality varies significantly. Japanese is tricky because it mixes three scripts (kanji, hiragana, katakana). Chinese simplified and traditional need different settings in some apps. Korean Hangul is actually quite OCR-friendly because of its block structure—I was surprised by how well it worked.
• Arabic, Hebrew, and other right-to-left scripts — The text direction can confuse some OCR engines. Letter connections in Arabic make individual character recognition harder. Results are improving but still less reliable than left-to-right scripts.
• Thai, Hindi, and other complex scripts — These have connecting marks, tone marks, and stacking behaviors that make OCR harder. App support varies a lot. Test before you travel.
• Handwritten text in any language — Always worse than printed. A handwritten menu on a chalkboard is going to give you rougher results than a printed paper menu.

If you know you’ll be dealing with a specific script, test your setup before you leave. Take a photo of some text in that language (find something online, display it on another screen, photograph it) and see what your OCR app produces. Better to discover a problem at home than at a pharmacy counter.

My setup for a language-barrier trip

Over several trips, I’ve settled into a prep checklist that takes about 15 minutes and covers most situations:

Before you leave:

• Install or update Google Translate and download every language you might encounter (not just the primary one—download neighboring countries’ languages too if you’re traveling overland)
• Make sure your OCR app is set to detect the right language for your destination
• Download the language in Apple Translate too, if you use iPhone—it’s a backup that works from the share sheet
• Optionally install DeepL and download offline data—its translations for European languages are often more natural than Google’s
• Test the whole flow: photograph some text in the target language, run OCR, paste into translator, check the result

At the restaurant or on the street:

• If you have data: Google Translate camera for the instant overlay. Point, read, order.
• If data is spotty or absent: take the photo, run it through an OCR app or Live Text to extract the text, paste into your translator with the downloaded language.
• For medication or anything safety-critical: always double-check the translation. Use two different translators if possible. If you’re still unsure, ask a local or a pharmacist. Machine translation is good, but it’s not perfect, and dosage information is not the place to guess.

Extra tips:

• Screenshot any important translations so you have them even if the app crashes or you close it
• If a waiter or local helps you understand a menu, write down what they said—you’ll see the same dishes elsewhere
• Save commonly needed phrases (allergies, dietary restrictions, hotel address) as plain text in your Notes app so you don’t have to translate them every time

When photo translation isn’t enough

I want to be honest about the limits. Photo translation works well for menus, signs, labels, and short text. It’s less reliable for:

• Long documents — Contracts, legal text, instruction manuals. OCR might handle it, but machine translation of complex sentences with specific terminology is often inaccurate. For anything you need to rely on, get a human translator.
• Handwritten signs — Chalkboard menus, handwritten notes from a host. OCR accuracy drops, and the translation quality drops with it because the input text has errors.
• Rare or regional languages — Some languages simply aren’t well-supported yet. If your destination uses a language with fewer than a few million speakers, check app support before you go.
• Cultural context — A menu item might translate literally to “grandmother’s pot” and that tells you nothing about what the dish actually is. Machine translation gives you words, not cultural knowledge. Sometimes asking someone is still the best option.

---

Translating text from a photo when you’re traveling is a solved problem if you set it up before you leave. Download your languages, have an OCR app ready, and know which method to use with and without data. The camera overlay is magic when you have a connection; the scan-and-translate flow is your safety net when you don’t. Spend 15 minutes prepping before the trip and you’ll have a secret weapon that turns illegible menus and mysterious signs into something you can actually read. For more on getting text out of images, see how I convert images to text on my iPhone and five ways to extract text from an image.