Recognition and translation are different steps
Character recognition identifies shapes, lines and reading order. Translation then interprets the resulting words in context. If the source character is read incorrectly, even a fluent translation may be wrong.
HonyakuMiru uses DeepL’s image translation service for translated images. The separate text extraction button reads the translated image using Cloudflare Workers AI. It does not return DeepL’s internal OCR text.
Improve the input before retrying
Increasing the size of a tiny image does not restore missing character details. Whenever possible, return to the original file or take a clearer photo.
- Use a sharp, straight image with good contrast and no glare.
- Crop a meaningful region if a long screenshot makes the text too small. Keep enough context to understand headings and labels.
- Compare names, digits and line order after translation or text extraction. Correct any copied text before using it.
Choose the output you need
Download an image when placement matters, such as a menu or diagram. Extract text when you need to search, copy or edit the words. Text extraction may flatten columns or choose a different reading order.
The extraction feature is a beta. It may omit text, join unrelated lines or misread technical words. Check the extracted text against the translated image, and then compare important meaning with the original source.
Official sources
Official feature descriptions checked on October 26, 2026. Service availability and terms can change.