toolgarden.xyz
中文

Image OCR (Text Recognition)

Free online OCR that extracts text from images, ideal for scans, screenshots, and photos

Upload image

Recognition settings

Recognition runs text detection, orientation checks, cropped recognition, and result merging in a local Worker. The first run downloads model assets.

Extracted text

Upload an image and click Start

About this tool

OCR analyzes letter shapes in an image and converts pixels into selectable text. The tool supports English, Simplified Chinese, Traditional Chinese, and Japanese, loading the selected language model and processing in the browser for screenshots, scans, and photographs.

Accuracy depends on sharpness, font, orientation, language, layout, and background interference. Tables, handwriting, vertical text, decorative type, and low-contrast photos commonly lose structure or confuse characters, so output must be checked against the image and must not be the sole source for amounts, contracts, or identity data.

How to use it

  1. Prepare a clear image

    Crop unrelated regions, correct orientation, and improve contrast between text and background where possible.

  2. Select the main language

    Choose the model matching most of the text, start recognition, and wait for processing to finish.

  3. Proofread every section

    Compare digits, punctuation, proper names, and line breaks against the source before further use.

Supported range and limits

Recognition engine
An open-source OCR engine; the first run downloads tens of megabytes of recognition data, cached thereafter
Recognises well
Scans, screenshots, printed type, and text that is high-contrast and level
Recognises poorly
Handwriting, decorative type, skewed or curved text, low-resolution photos, and text over busy backgrounds
Layout
The output is text; multi-column layouts, tables and wrapped images do not survive
Improving accuracy
Straighten with Image Rotate, crop away irrelevant areas, and upscale if needed; this beats running OCR on the raw photo
Where it runs
Recognition data is downloaded, but the image itself is processed in the browser and never uploaded

When you would use it

  • Extracting screenshot text

    Copy content from an interface, error image, or caption where text selection is unavailable.

  • Transcribing scanned material

    Create a searchable draft from printed pages, then restore paragraphs and table structure manually.

  • Getting editable text back out of a screenshot

    When only the screenshot survives, recognising and proofreading is far quicker than retyping from the screen.

What to know before you start

  • Selecting a language for the first time may download a model, with time depending on the network and device.
  • OCR can confuse similar glyphs such as `0` and `O` or `1` and `l`; verify critical identifiers character by character.
  • Images containing personal or confidential information remain subject to organizational handling and retention rules.

Related concepts

OCR
Optical Character Recognition, the process of inferring text characters from shapes in an image.
language model
Recognition data describing characters and word forms for a particular writing system.

Frequently asked questions

Does image OCR upload my picture?
No. Text recognition runs locally in your browser, so screenshots and photos are never uploaded to a server.
Does it recognise Chinese text?
Yes. It reads languages such as Chinese and English, making it useful for text in screenshots, scans and photos.
What affects recognition accuracy?
Sharp, high-contrast, well-aligned images work best; blurry or handwritten content lowers accuracy.
How do I improve poor recognition?
Try these in order: straighten the text with Image Rotate, crop away everything that is not text, and upscale if the resolution is low. Those three usually help more than switching tools.
Does it read handwriting?
Poorly. The engine is trained mainly on printed type, and handwriting varies far beyond what it handles. Neat handwriting may yield fragments; anything hurried will not be usable.