🔍 OCR 图片文字识别（中英文）OCR — Image to Text

使用说明与常见问题How-to & FAQ

「OCR 图片文字识别（中英文）」是什么？What is the OCR — Image to Text tool?

在浏览器里识别图片和扫描 PDF 中的文字（OCR）：支持简体中文和英文混排，可拖放、选择或直接 Ctrl+V 粘贴截图，自动放大小图提高准确率，显示置信度，结果可复制或下载 TXT；图片不上传服务器。Recognise text in images and scanned PDFs (OCR) right in your browser — Simplified Chinese and English, mixed. Drop, choose or just Ctrl+V paste a screenshot; small images are upscaled for accuracy, confidence is shown, and the result can be copied or saved as TXT. Images are never uploaded.

如何使用「OCR 图片文字识别（中英文）」？How do I use the OCR — Image to Text tool?

在「OCR 图片文字识别（中英文）」中，按需设置「识别语言」、「版面」等选项，然后点击「开始识别」、「复制」、「下载 TXT」按钮即可。支持拖放文件。In the OCR — Image to Text: Set "Language", "Layout", then click "Recognise", "Copy", "Download TXT". Drag-and-drop files is supported.

识别结果里中文和英文/数字之间的空格会保留吗？Are spaces between Chinese and Latin text kept?

勾选“去掉汉字之间的空格”时，只删除两个中文字符之间被引擎误加的空格；中文与英文、数字之间的空格按引擎输出保留，因为无法可靠判断原文是否有空格，如有需要请手动校对。With "Remove spaces between Chinese characters" ticked, only spaces the engine inserts between two Chinese characters are removed. Spaces between Chinese and Latin letters or digits are kept as the engine outputs them, because it cannot reliably tell whether the original had a gap — proofread if it matters.

识别准确率怎么样？How accurate is it?

使用 Tesseract OCR 引擎（LSTM 神经网络模型）。清晰的截图、扫描件和印刷体通常能达到很高的准确率；手写体、艺术字、模糊或倾斜严重的照片效果会明显下降。结果会显示平均置信度，低于 60% 时建议校对。It uses the Tesseract OCR engine (LSTM neural-network models). Clear screenshots, scans and printed text usually come out very accurately; handwriting, decorative fonts and blurry or heavily skewed photos are much weaker. The mean confidence is shown — proofread anything below 60%.

为什么第一次识别要等一会儿？Why does the first recognition take a while?

第一次需要下载识别引擎（约 4 MB）和语言模型（英文约 3 MB、简体中文约 1.7 MB），之后浏览器会缓存，再次识别会快很多。所有计算都在你的设备上进行，图片不会上传。The first run downloads the engine (~4 MB) and language models (English ~3 MB, Simplified Chinese ~1.7 MB); the browser caches them, so later runs are much faster. All processing happens on your device and images are never uploaded.

支持繁体中文、日文或手写吗？Does it support Traditional Chinese, Japanese or handwriting?

目前只提供简体中文和英文模型。繁体字可能被识别成相近的简体字或识别失败，日文、韩文和手写体不支持。Only Simplified Chinese and English models are included. Traditional characters may be read as similar simplified ones or missed, and Japanese, Korean and handwriting are not supported.

PDF 怎么识别？How do PDFs work?

扫描版 PDF 会把每页渲染成图片后识别（最多前 20 页）。如果 PDF 本身就有可选中的文字，用“PDF 转文本”工具直接提取会更快、更准确。Scanned PDFs are rendered page by page and recognised (first 20 pages). If the PDF already has selectable text, the PDF to Text tool extracts it faster and exactly.

怎样提高识别效果？How can I improve results?

尽量使用清晰、正对、光线均匀的图片；只需要一行文字时把“版面”设为“单行文字”；深色背景浅色字请勾选“反色”；选择与图片一致的语言（纯英文时选“英文”更准）。Use sharp, straight, evenly lit images; set Layout to "Single line" for one line of text; tick Invert for light text on a dark background; and pick the matching language (English-only text is more accurate with "English").

图片会被上传到服务器吗？Are my images uploaded?

不会。「OCR 图片文字识别（中英文）」使用浏览器 Canvas 在本地处理图片，处理后的文件直接从你的浏览器下载。No. The OCR — Image to Text processes images locally with the browser Canvas API and the result downloads straight from your browser.

「OCR 图片文字识别（中英文）」免费吗？需要注册吗？Is the OCR — Image to Text free? Do I need an account?

「OCR 图片文字识别（中英文）」完全免费，无需注册或安装，打开网页即可使用，手机和电脑都能用；输入的内容只在你的浏览器本地处理。The OCR — Image to Text is completely free with no sign-up or install and works on phones and desktops; what you enter is processed only in your browser.