Chinese image to text is for the moments when the text you need is stuck in a picture: a WeChat screenshot with an address in it, a supplier’s spec sheet sent as an image, a menu, a slide from a lecture, a page of a textbook you’re studying, or subtitles paused on a video. Paste or drop the image in and copy the characters out.
Chinese and English are both selected on this page, since so much Chinese text includes English brand names, units, model numbers and email addresses. Both are read in the same pass.
Simplified or Traditional
Image to Text App has two separate Chinese models: Chinese (Simplified) and Chinese (Traditional). Simplified characters are used in mainland China and Singapore. Traditional characters are used in Taiwan, Hong Kong and Macau, and in many older books.
Many common characters differ between the two: 门 and 門, 书 and 書, 说 and 說, 国 and 國, 学 and 學. Read Traditional text with the Simplified model and those characters come out wrong or not at all. If a result has far more underlined characters than you’d expect, the wrong set is the first thing to check.
To switch, open the language menu (the globe), untick Chinese (Simplified) and tick Chinese (Traditional). When you close the menu, the page is read again with the new selection. If a page mixes both sets, you can select both, though reading gets slower with each language you add.
The engine reads characters as they are printed. It doesn’t convert Simplified to Traditional or the other way around, so you get exactly the characters in your image.
No spaces between words
Chinese is written without spaces, but OCR engines work in units they call words, and Image to Text App puts a space between those units. As a result, you may see spaces between some characters where the original has none. Joining wrapped lines into paragraphs adds a space at each join too, which looks odd in Chinese.
Both are quick to clean up with find and replace (Ctrl+F, or ⌘F on a Mac):
- Text with no English: search for a single space and replace all with nothing.
- Text with English in it: tick regexp, search for
(?<=\p{Script=Han}) (?=\p{Script=Han})and replace all with nothing. This removes only the spaces between two Chinese characters and leaves the spaces between English words alone.
Characters that look alike
A common mistake in Chinese OCR is a character swapped for its near twin. These are worth checking first:
- 己, 已 and 巳, which differ only in how far one stroke closes
- 未 and 末, where the length of the top stroke decides which is which
- 土 and 士, 人 and 入, 日 and 曰
- 一 (one) and a dash, and 〇, the letter O and the digit 0
- Full-width punctuation such as ,。、;:?! and quotation marks like 「」 or “”, which can be dropped or swapped for their narrow Western versions
Words the engine was unsure about are underlined, with a button to step from one to the next. Click any line and its place lights up on the image so you can compare the strokes.
Vertical text and traditional layouts
The Chinese options here read horizontal lines. Text set vertically, in columns from top to bottom, is common in Traditional Chinese books, some newspapers in Taiwan and Hong Kong, couplets and shop signs, and it isn’t something on-device reading handles well. It usually comes out wrong or in the wrong order, and there’s no vertical option in the language list.
If a page mixes the two, crop the horizontal parts and read those. For a few lines of vertical text, the text recognition built into your phone, such as Google Lens or Live Text on an iPhone, is worth trying.
Getting a cleaner result
A Chinese character can have more than twenty strokes in the space an English letter takes up, so small text loses detail quickly.
- Zoom in before you screenshot. More pixels per character is the single biggest help, especially for dense characters.
- Move closer for photos. Digital zoom only enlarges pixels; stepping closer gives real detail. See photo to text for more on photographing pages.
- Crop to the text you need. Leave out stickers, icons and decorative headings, which the engine would otherwise try to read.
- Straighten angled shots of menus, signs and pages with four-corner straightening.
Dark-mode screenshots need nothing special: Auto improve inverts light text on a dark background and lists each step it took. For more, read how to get accurate OCR results.
Where the text goes next
The result is ordinary Unicode Chinese, so it pastes into WeChat, email, Google Docs, Word or a translator like any typed text. You can also download it:
- Word (.docx) for editing. See image to Word.
- Searchable PDF, which keeps your original image and adds an invisible text layer, so you can search the Chinese with Ctrl+F in a PDF reader.
- PDF, which for Chinese opens your browser’s print dialog, where you choose “Save as PDF” so every character is drawn with a proper font.
- Plain text, Markdown or HTML for notes apps and websites.
For a series of screenshots, such as a long chat or a multi-page document, each image becomes a page. Drag them into order and use Whole document view to copy or download everything together.
Private by default
Screenshots of chats, invoices and documents often contain names, addresses and account details. Image to Text App reads them in your browser, on your device, and nothing is uploaded. Read more in is online OCR private?
If your image is in Japanese rather than Chinese, use Japanese image to text instead, since kana need the Japanese model.