Markdown AI Rewriter

What it solves

Existing Markdown drafts often need polishing, rewriting, style changes, or terminology cleanup. Direct LLM rewrites can easily break heading hierarchy, code blocks, tables, and image placement. Batch work also needs to balance provider availability, token cost, output stability, and downstream publishing steps. Markdown AI Rewriter focuses on improving an existing document while preserving Markdown structure as much as possible.

How it works

It processes text in section or full mode. section splits the document by headings and calls the model chapter by chapter, which suits long technical articles, blog posts, and newsletters with controllable cost and speed. full sends the whole document at once, producing a more coherent tone but using more tokens on long files. It can use OpenAI, Anthropic, Gemini, DeepSeek, Doubao, MiniMax, and other providers, switching via -p and API keys. Beyond rewriting, it can explicitly enable image, video, and music generation for captions, demo clips, or background audio. Retention of lists and quotes depends on model output; if strict structure matters, state in the prompt to preserve tables, code blocks, and lists.

Limits

It is best for existing Markdown, not for writing from scratch, pure MD-to-HTML conversion, or structure analysis without rewriting. Rewrite, image, video, and music features are disabled by default and must be enabled explicitly. Output structure is not guaranteed lossless; test small samples first to verify headings, code, tables, and image positions before connecting it to a cleaning or publishing pipeline.