Text recognition, translation and image rendering
A typical workflow first finds text areas, recognizes the characters and translates the recognized text. A rendering step then fits the translation back into the image. An error at any stage can appear in the final result.
HonyakuMiru sends the prepared image to DeepL’s image translation service. Its optional text extraction reads the translated image separately using Cloudflare Workers AI; extracting text is not required to download an image.
Translate an image in three steps
Use the highest-quality original available. Avoid repeatedly saving a screenshot as JPEG, which can blur small characters.
- Choose or paste your image into the tool. Check the preview and source language.
- Choose a target language and start translation. Keep the page open while it runs.
- Compare the original and result, particularly names, numbers and labels. Download PNG or WebP when ready.
Which images need more checking?
Clear, straight, well-lit printed text is easier to review than handwriting, decorative lettering or overlapping text. Long screenshots and dense flowcharts may make small words difficult to recognize after resizing.
Translations can be longer than the source. Line breaks, font size and background details may change. If a region is missed, crop a readable section and translate that separately. Check the original whenever a decision depends on the text.
Official sources
Official feature descriptions checked on October 26, 2026. Service availability and terms can change.