Multilingual Bidirectional Translation Assistant

Problem to Solve

In multilingual documents, design specs, and image-based copy, the hard part is not only translating words but also choosing the right direction, keeping terminology consistent, extracting text from images, and preserving long-text structure. Manually switching languages, cleaning OCR output, and checking domain terms adds unnecessary overhead.

How It Works

The skill infers translation direction from input: Chinese text is translated into English; English, French, Russian, German, Spanish, Arabic, Japanese, Korean, and other supported languages are translated into Chinese. For text input, it outputs the translation directly without extra labels. For images, it first uses OCR to extract visible text, then translates it while preserving visual hierarchy such as headings, body copy, and annotations.

It also handles art and design terminology with attention to common industry usage, including terms like UI/UX, kerning, leading, PBR, and CMYK. Proper nouns such as tool names, font names, and standards are retained. When issues occur, it surfaces likely causes such as network failure, poor image quality, unrecognized language, or possible terminology drift instead of failing silently.

Boundaries

It is suitable for short sentences, paragraphs, documents, design copy, and image text translation, but it does not guarantee readable results from every image. Low-resolution, garbled, text-less, or unsupported images may require re-uploading. Very long documents may be processed in batches, and key terms, brand names, and business wording should still be reviewed by a human.