File Translate — Layout-Preserving Immersive Translation

Translate documents, images, and text while keeping the original look and feel.

Unlike plain-text translation, File Translate understands your document's layout — tables, headers, charts, embedded images, and formatting are all preserved. Images get in-place text replacement that blends with the scene. The result reads like it was written in the target language from the start.

✨ Highlights

📄 Document translation — PDF, Word, PPT, Excel, ePub, HTML, and more → translated PDF with original layout intact
🖼️ Image translation — scene-aware text replacement, not ugly overlays. Fonts, colors, and backgrounds blend naturally
📝 Text translation — fast batch translation for strings and paragraphs, 100+ languages
🔀 Smart routing — Word files automatically use the deepdoc v2 engine for best fidelity; other formats route to v1. Zero config needed
🌍 100+ languages — auto-detects source language

Prerequisites

Requires a 360 AI Platform API Key.

Create a key at 360 AI Platform
（新注册用户会获赠 50 元体验金，可直接用于翻译服务）

Set the environment variable in your shell config:

export TRANSLATE_360_API_KEY="your-api-key-here"

source ~/.zshrc to reload

Security tip: Use a dedicated API key with minimal billing scope for this skill. Rotate or revoke it when you stop using the skill.

No pip install — pure Python 3 stdlib.

Document Translation

python3 scripts/translate.py doc --tl <target_lang> --file <doc_path> [--out <output_dir>] [--timeout 3600]

Supported: pdf, doc, docx, xls, xlsx, ppt, pptx, csv, epub, rtf, html, xhtml, txt
Not supported: md, mobi
Word files (.doc/.docx) → deepdoc v2 engine (best quality)
All other formats → v1 engine
Routing is automatic and transparent
Max 100MB per file. Output is always PDF
Default timeout: 3600s (1 hour — large docs can take a while)

Image Translation

python3 scripts/translate.py image --tl <target_lang> --file <image_path> [--out <output_path>]
python3 scripts/translate.py image --tl <target_lang> --url <image_url> [--out <output_path>]

Scene-aware: replaces text in-place matching original font style, color, and background
Supported: JPG, PNG, WEBP, BMP, GIF (first frame). Max 10MB
Target languages: zh, en, ja, fr, ru, ko, pt, th, es, it, hi, pl, vi, id
Without --out, prints translated image URL (valid 6 months)

Text Translation

python3 scripts/translate.py text --tl <target_lang> [--sl <source_lang>] "text1" "text2" ...

Max 50 texts per call, source auto-detected if omitted
Common codes: zh (Chinese), en (English), ja (Japanese), ko (Korean), fr, de, es, ru

Quick Reference

| What you have | Command | What you get | |---|---|---| | A Word doc | doc --tl en --file report.docx | Translated PDF, layout preserved | | A PDF manual | doc --tl zh --file manual.pdf | Translated PDF | | A slide deck | doc --tl ja --file slides.pptx | Translated PDF | | A photo with text | image --tl en --file sign.jpg | New image, text replaced in-scene | | A paragraph | text --tl zh "Hello world" | Translated string |

Privacy & Data Handling

This skill uploads user-provided content to the 360 AI Translation API for processing. Read this section carefully before use.

What is uploaded

Text translation: Plain text strings are sent to api.360.cn via HTTPS.
Image translation: Images are sent as base64-encoded data or via URL to api.360.cn via HTTPS.
Document translation: Entire document files (PDF, Word, PPT, Excel, etc.) are uploaded to api.360.cn (v1 engine) or api.360.cn/deepdoc (v2 engine for Word files) via HTTPS multipart upload.

Data retention & access control

Translated image URLs: Returned by 360's CDN and typically expire after 3 months. These URLs are not authenticated — anyone with the link can access the image during the validity period. Download results promptly and avoid sharing URLs containing sensitive content.
Uploaded documents: Document processing is asynchronous. The translated PDF is available via a temporary download URL returned by the API. Refer to 360 AI Platform terms for the provider's data retention, deletion, and access control policies.
No local persistence: This skill does not persist uploaded files or translated results on disk beyond the current execution. Output files are only saved when you explicitly pass --out.

Credentials

The only credential required is TRANSLATE_360_API_KEY, used solely to authenticate with the 360 translation endpoints.
The skill does not collect, log, or transmit API keys or any data beyond what is required for the translation request.

Recommendations

Do not use this skill for highly sensitive, confidential, or classified documents unless you accept that content will be transmitted to and processed by a third-party service (360 AI, operated by 360 Group).
Use a dedicated API key with minimal privileges. Rotate or revoke it when you stop using the skill.
Test with non-sensitive sample files first before processing important documents.
Review the 360 AI Platform privacy policy and terms of service for full details on data handling, retention, and sharing.

API Details

Full request/response schemas, error codes, and language list: references/api-reference.md

Pricing

Text: ¥30 / 1M characters
Image: ¥0.04 / image
Document: ¥0.2 / page