Markdown Translator#

Markdown is the text format I use most frequently in my daily work, but existing translation tools often struggle to preserve its original formatting. That's why I developed md-translator, a translation tool optimized specifically for Markdown. It is designed to accurately translate content while fully preserving the original format.

Supported Markdown Elements#

md-translator currently supports parsing the following Markdown syntax and retains their formatting during translation:

• FrontMatter metadata (---)
• Headings (#)
• Links ([text](url))
• Unordered lists (- / * / +)
• Ordered lists (1. 2. 3.)
• Emphasis (**bold**, _italic_)
• Code blocks (```)
• Inline code (`code`)
• Blockquotes (> quote)

Additionally, md-translator can extract plain text content, stripping away Markdown syntax and optionally hiding elements like links and code blocks, making it suitable for other purposes.

Translation API#

This tool supports 5 translation APIs and 5 LLM (large language model) interfaces, allowing users to choose the appropriate translation method based on their needs:

Comparison of Translation APIs#

• DeepL: Ideal for long texts with smoother and more natural translations; however, it does not support web API calls and requires local or server-side proxy usage.
• Google Translate: Offers stable translation quality, suitable for short sentences and UI text, and supports web API calls.
• Azure Translate: Provides the widest range of language support, meeting diverse multilingual translation needs.
• GTX API/Web: A free translation option suitable for small-scale use, though its stability is average.

For higher translation speed and quality, you can apply for an API Key from Google Translate, Azure Translate, or DeepL Translate. Refer to the related API application tutorial for the application process.

LLM Translation (AI Large Models)#

This tool provides access to five mainstream AI large models (LLMs) or interfaces: OpenAI, DeepSeek, Siliconflow, Groq, and Custom LLM.

• Use Cases: Ideal for handling content that requires high-level language understanding, such as literary works, technical documents, and multilingual materials.
• Customizability: Supports custom system prompts and user prompts, allowing flexible adjustments to translation style and terminology preferences to better suit different scenarios.
• Temperature Parameter: Controls the randomness of translation results. Higher values produce more creative outputs but may affect consistency and accuracy.

The Custom LLM option allows integration with third-party services or local inference platforms (such as ollama) by configuring the API endpoint and model name. For example, the default API endpoint for a local ollama setup is:

http://127.0.0.1:11434/v1/chat/completions

The default model used is llama3.2. For LM Studio, the local API endpoint is:

http://localhost:61234/v1/chat/completions

To achieve better translation quality, it is recommended to use qwen2.5-14b-instruct or a higher-performing model in the Custom LLM setup.

Language Support#

In terms of multilingual translation capabilities, Google, Azure, and large language models (LLMs) support translation between hundreds of languages, while DeepL currently supports only 30 mainstream languages. Therefore, in terms of language coverage, DeepL is relatively weaker.

For a detailed list of supported languages, please refer to the official documentation:

• Languages supported by DeepL

• Languages supported by Google Translate

• Languages supported by Azure Translator

API Parameters#

Chunk Translation Size#

For text files with contextual relationships—such as subtitles or Markdown documents—this tool automatically merges multiple lines into "chunks" for translation. The chunk size refers to the maximum number of characters per grouped block. The character limits for each translation service are as follows:

• DeepL API: Up to 128K characters per request
• DeepLX Free: Up to 1,000 characters per request
• Azure Translate: Up to 10,000 characters per request
• Google Translate:
  • Web version: Up to 5,000 characters per translation
  • Cloud Translation API: Up to 30,000 characters per request

Note: Google Translate disrupts line breaks during processing, so chunked translation is not used with this service.

Delay Time#

Delay time sets the wait interval between chunk translations. When processing large volumes of text, some translation APIs may respond slowly—especially under poor network conditions or when using free interfaces. In such cases, delay settings are particularly important.

For example, when testing with Azure Translate’s free tier, it is recommended to set the delay time to 5,000 milliseconds or more to avoid empty responses or errors.

Translation Rate#

Setting the translation rate too high may result in empty API responses or cause requests to be flagged as abnormal. It's recommended to adjust the rate based on the specific translation service and its stability to improve success rates and maintain reliable performance.

Feature Description#

Translation Cache#

This tool introduces an optional local translation cache to improve translation efficiency and reduce resource consumption:

• Cache rules: Each translation result is stored with a unique key formatted as source text_target language_source language_translation API_model settings.
• Cache hit condition: The local cache result is used only when the parameters match exactly, ensuring accuracy.
• Cache purpose: Avoid repeated translations, reduce API calls, and improve translation speed.

To disable the use of translation cache, you can uncheck "Use translation cache" or click "Clear translation cache" in the API settings.

Multilingual Translation#

Supports translating the same file into multiple languages at once, which is especially suitable for international video content:

• For example: Translate an English file simultaneously into Chinese, Japanese, German, and French for the convenience of global users.
• Supports 35 mainstream languages and will continue to expand.

Usage Notice#

When using this tool, please note the following:

• DeepL support: Since the DeepL API does not support direct calls from the web, a server-side forwarding interface is provided solely for data transmission, and it will not collect user data. For better stability, users can also choose to deploy this interface themselves.
• Using the DeepLX free interface may sometimes return null. Please wait a moment and try again, or use your own API KEY or deploy your own forwarding interface.
• API Key security: This tool does not store your API key; all configuration data is saved in your local browser.
• GTX Web interface: This interface places considerable load on the server, so it is recommended to enable it manually only when deploying locally. Please avoid using it in networks with a global proxy enabled to prevent translation errors.