Agent-Skills.md

Agent Skills: PDF to Audio Converter (MiniMax)

Convert PDF files to MP3 audio using MiniMax MCP Server's text-to-audio tool. Use when user wants to convert a PDF to audio/MP3, create audiobook from PDF, or text-to-speech for PDF documents. Requires PDF file path; voice ID is optional (auto-selects based on content).

UncategorizedID: aicoder2048/story-writer/pdf2audio-minimax

Repository

aicoder2048/story-writer
aicoder2048
12

Install this agent skill to your local

pnpm dlx add-skill https://github.com/aicoder2048/story-writer/tree/HEAD/.claude/skills/pdf2audio-minimax

Skill Files

Browse the full folder contents for pdf2audio-minimax.

Download Skill

Loading file tree…

.claude/skills/pdf2audio-minimax/SKILL.md

Skill Metadata

Name
pdf2audio-minimax
Description
"Convert PDF files to MP3 audio using MiniMax MCP Server's text-to-audio tool. Use when user wants to convert a PDF to audio/MP3, create audiobook from PDF, or text-to-speech for PDF documents. Requires PDF file path; voice ID is optional (auto-selects based on content)."

PDF to Audio Converter (MiniMax)

Convert PDF documents to MP3 audio files using MiniMax text-to-audio.

Input Format

/pdf2audio-minimax <pdf_file_path> [voice_id]

Examples:

  • /pdf2audio-minimax /path/to/story.pdf (auto-select voice)
  • /pdf2audio-minimax /path/to/story.pdf Chinese (Mandarin)_Gentleman

Workflow

1. Parse Arguments

Extract from user input:

  • pdf_path: Full path to the PDF file (required)
  • voice_id: Voice identifier (optional)

2. Read PDF Content

Read: file_path = <pdf_path>

3. Extract Text & Metadata

Extract readable text, skipping page numbers and headers. Keep chapter titles, body text, dialogue.

Extract metadata for file naming:

  1. Story Name: Extract from the PDF's parent directory name (e.g., /path/我的世界/chapters/file.pdf → 我的世界)
  2. Chapter Number: Extract from filename pattern *-XX.pdf or content like 第X章 (e.g., 我的世界-03.pdf → 03)
  3. Chapter Title: Extract from the first chapter heading in content (e.g., 第3章：迷宫里的真心话 → 迷宫里的真心话)

4. Select Voice

If voice_id provided: Use the specified voice.

If voice_id not provided: Auto-select based on content keywords:

| Content Keywords | Voice ID | |-----------------|----------| | 童话、儿童、小朋友、冒险 | Chinese (Mandarin)_Cute_Spirit | | 言情、爱情、浪漫、甜蜜 | Chinese (Mandarin)_Sweet_Lady | | 科幻、悬疑、历史、战争 | Chinese (Mandarin)_Gentleman | | 新闻、报告、公告 | Chinese (Mandarin)_News_Anchor | | Default | Chinese (Mandarin)_Gentleman |

For complete voice list, see references/voices.md.

5. Convert to Audio

mcp__MiniMax__text_to_audio:
  text: <extracted_text>
  voice_id: <selected_voice_id>
  output_directory: <audiobook subdirectory in story folder>
  language_boost: "Chinese"
  speed: 1

6. Rename Output File

After MiniMax generates the MP3, rename it to follow the naming convention:

<STORY_NAME>_<CHAPTER_NUMBER>_<CHAPTER_TITLE>.mp3

Examples:

  • 我的世界_03_迷宫里的真心话.mp3
  • 文明的最后一个变量_01_第一次接触.mp3

Rename command:

mv "<generated_file>.mp3" "<output_directory>/<STORY_NAME>_<CHAPTER_NUMBER>_<CHAPTER_TITLE>.mp3"

7. Output

Report the final MP3 path with the properly formatted filename.

Quick Reference: Common Voices

| Use Case | Voice ID | |----------|----------| | Audiobook (male) | Chinese (Mandarin)_Gentleman | | Audiobook (female) | Chinese (Mandarin)_Soft_Girl | | Children's story | Chinese (Mandarin)_Cute_Spirit | | News/Formal | Chinese (Mandarin)_News_Anchor |

Full voice reference: references/voices.md