# Convert To Markdown

> Fetch web content. Use when the user asks to extract text from PDFs, Word docs, PowerPoint, Excel, images, audio, EPUBs, MHTML, YouTube transcripts, or any URL — even if they say 'read this file', 'get the text from', 'extract content', or 'transcribe'.

- Skill: `varlabz/convert-to-markdown` (Agent Skill, multi-file: 10 files)
- Install (CLI): `npx skillmds@latest add varlabz/convert-to-markdown`
- Raw SKILL.md: https://api.skillmd.com/api/skills/varlabz/convert-to-markdown/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Data & Analytics
- Author: varlabz (https://skillmd.com/u/varlabz)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/varlabz/convert-to-markdown

---


# Convert Files and URLs to Markdown

All scripts live in `scripts/`. Run them "as is".

## Tool Selection

| Input type | Command | Script location |
|------------|---------|-----------------|
| `.pdf`, `.docx`, `.pptx`, `.xlsx`, `.xls`, `.csv`, `.json`, `.xml`, `.epub`, images (`.jpg`/`.png`/etc.), audio (`.mp3`/`.wav`/etc.), `.html`, `.txt`, `.md` | `uvx markitdown <path>` | built-in CLI |
| YouTube URL or video ID | `./scripts/youtube-transcript <url_or_id>` | `scripts/youtube-transcript` |
| Regular web URL (Crawl4AI server running) | `./scripts/web-to-md <url>` | `scripts/web-to-md` |
| `.mhtml` file or MHTML URL | `./scripts/mhtml-to-md <source>` | `scripts/mhtml-to-md` |

## YouTube Transcript

```bash
./scripts/youtube-transcript <url_or_video_id> [--info] [--lang <code>] [--raw]
```

- `--info`: Include video metadata (title, channel, duration) before transcript.
- `--lang <code>`: Subtitle language (e.g. `en`, `de`). Default: auto-detect source language.
- `--raw`: Output raw VTT instead of cleaned text.

## MHTML to Markdown

```bash
./scripts/mhtml-to-md <path_or_url>
```

Accepts local `.mhtml` files or remote URLs serving MHTML content.

## Web Page to Markdown (via Crawl4AI)

```bash
./scripts/web-to-md <url> [-o output.md] [-s server_url]
```

Requires a [Crawl4AI](https://github.com/unclecode/crawl4ai) server running on `http://localhost:11235` (or set `$CRAWL4AI_URL`). The server fetches the full page HTML; this script converts it to Markdown.

**No Crawl4AI?** Fall back:
```bash
curl -o page.html <url> && uvx markitdown page.html
```

## MarkItDown (General File Conversion)

All other formats use `uvx markitdown`:

```bash
uvx markitdown <path> [-o output.md]
```

Supports: PDF, DOCX, PPTX, XLSX/XLS, images (EXIF + OCR), audio (transcription), EPUB, CSV, JSON, XML, HTML, TXT, ZIP.

## When Things Go Wrong

| Issue | What to do |
|-------|-----------|
| Scanned PDF yields no text | Try `uvx markitdown <path>` treating it as an image — markitdown runs OCR on images |
| Password-protected file | Remove password first; markitdown can't decrypt |
| Large ZIP archive | Extract specific files first, then convert individually |
| YouTube age-restricted video | yt-dlp will fail — no workaround |
| Audio transcription poor quality | Depends on audio clarity; nothing we can do from here |
| Excel has many sheets | All sheets included as separate tables in output |