OCR-Resistant Image Obfuscator (Make Text Harder to Read for OCR)
Choose an image and adjust "Noise", "Distortion", and "Line overlay" to keep it readable to people while making automatic reading by OCR (text recognition) more difficult. It can help reduce the chance of your own images being scraped and read in bulk. To hide only part of an image, try the Pixelate, Mosaic & Color Reduction tool; to add text, use Add Text or Watermark to Image tool.
🔒 Runs in your browser · Your data is never sent to a server
Important: this processing cannot be guaranteed to stop OCR or AI image recognition from reading your text. As the technology advances, processed text may still be read. Do not rely on it to protect passwords, personal data, or other sensitive information; keep anything you truly need to hide out of the image.
You can also drag and drop an image file here
How to use
- Choose an image file (drag and drop also works).
- Pick a preset, or adjust the "Noise", "Distortion", and "Line overlay" sliders to set the strength.
- Zoom in on the before and after previews and check that it is still readable to the eye. Press "Regenerate with a new pattern" to change the pattern.
- When it looks right, save it with "Download as PNG".
Notes
- OCR engines and AI image recognition keep improving, so this processing cannot be guaranteed to stop OCR from reading your text. Do not rely on it to protect sensitive information.
- Higher strength tends to make text harder for OCR to read, but also harder for people. Always zoom in on the preview and balance the strength against readability.
- Images with added noise, distortion, and lines can be harder to read for people who use screen readers (which rely on alt text) or magnification. For important content, also provide it another way, such as alt text.
- For animated GIFs, only the first frame is processed.
- The output is a lossless PNG. Because it contains noise, the file may be larger than the original.
- This tool is intended to help protect images you own from scraping and bulk collection. Do not use it on other people’s images or for improper purposes such as bypassing CAPTCHAs.
- All processing happens in your browser; the image you choose is never sent to a server.
Glossary
OCR (optical character recognition)
Technology that reads text in an image and converts it to text data. It is used to transcribe screenshots and photos, and in automated collection (scraping).
Adversarial noise
Noise or distortion that is barely noticeable to people but disturbs a machine-learning model’s judgment. This tool is similar in spirit, but it is not tuned to any specific OCR engine and its effect cannot be guaranteed.
Scraping
Collecting website content such as images and text automatically with a program. Text inside images can sometimes be extracted with OCR.
FAQ
Does this reliably stop OCR from reading my text?
No. OCR engines and AI image recognition keep improving, and processed text may still be read. Keep passwords, personal data, and similar information out of images in the first place.
What strength should I use?
Start with "Standard", zoom in on the processed preview, and check that it is still readable to the eye. Higher strength tends to make text harder for OCR to read, but also harder for people.
Does adding noise and distortion affect screen reader users?
Text in images reaches screen reader users through alt text, and people who read with magnification may find noisy text harder to read. For important content, also provide it another way, such as alt text.
Why is the output PNG only?
Lossy compression in JPEG or WebP can smooth out the added noise and lines and weaken the effect, so the result is saved as lossless PNG. Because of the noise, the file may be larger than the original.
Share this page