Output Generation Commands for Audio, PPT, and Mind-Maps in qiaomu-anything-to-notebooklm

The qiaomu-anything-to-notebooklm project recognizes six specific Chinese natural-language trigger phrases that function as output generation commands for audio, PPT, and mind-maps, routing requests to the corresponding NotebookLM generation endpoints.

The joeseesun/qiaomu-anything-to-notebooklm repository transforms URLs, PDFs, and text files into structured NotebookLM outputs through a command-line interface. By appending specific natural-language instructions to the CLI command, users can directly trigger the creation of audio podcasts, PowerPoint presentations, or hierarchical mind-maps without manual configuration. Understanding these output generation commands is essential for automating content conversion workflows.

Natural-Language Trigger Phrases for Each Format

The system maps specific Chinese phrases to output formats as documented in README.md (lines 31–34).

Audio and Podcast Generation

To generate an audio podcast file, use either "生成播客" or "做成音频". These phrases trigger the notebooklm generate audio command internally, producing an MP3 file (e.g., /tmp/source_podcast.mp3).

PowerPoint and Slide Generation

For presentation output, the trigger phrases "做成PPT" or "生成幻灯片" instruct the system to invoke notebooklm generate ppt. This generates a PDF-formatted slide deck (e.g., /tmp/source_presentation.pdf).

Mind-Map Generation

Mind-maps require "画个思维导图" or "生成脑图", which activate the notebooklm generate mind-map endpoint. The output is delivered as a JSON file representing the hierarchical node structure (e.g., /tmp/source_mindmap.json).

Command Parsing Implementation

The detection logic resides in main.py within the detect_input_type and deep_analysis functions. When you execute a command, the script:

• Parses the natural-language request to identify trigger phrases
• Uploads the source content to NotebookLM
• Invokes the specific generation endpoint matching the detected phrase

According to the source code, only these documented phrases trigger format-specific generation; other phrasing defaults to generic processing.

Practical CLI Usage Examples

Use these commands exactly as shown to generate specific output types:

# Generate audio podcast from web article

python main.py https://example.com/article "生成播客"

# Output: /tmp/article_podcast.mp3

# Create PowerPoint from PDF document

python main.py ./slides.pdf "做成PPT"

# Output: /tmp/slides_presentation.pdf

# Generate mind-map from social media post

python main.py https://x.com/user/status/123 "画个思维导图"

# Output: /tmp/tweet_mindmap.json

Summary

• The repository supports three distinct output formats: audio podcasts, PowerPoint presentations, and mind-maps
• Each format requires specific Chinese natural-language trigger phrases documented in README.md (lines 31–34) and SKILL.md
• The main.py entry point handles request parsing through the detect_input_type and deep_analysis functions
• Commands map directly to NotebookLM generation endpoints: notebooklm generate audio, notebooklm generate ppt, and notebooklm generate mind-map

Frequently Asked Questions

What are the exact trigger phrases for audio generation?

The system recognizes "生成播客" and "做成音频" as valid commands for audio output. These phrases are hardcoded in the detection logic within main.py and formally documented in README.md (lines 31–33).

Can I customize the output file paths?

The analysis indicates default temporary paths like /tmp/article_podcast.mp3, but the repository's main.py likely provides configuration options or uses the system's temporary directory. Review the implementation details in the deep_analysis function for path customization parameters.

Where else are these output generation commands documented?

In addition to README.md (lines 31–34), the trigger phrases are listed in SKILL.md for Claude Code integration, ensuring consistent command recognition across different interfaces when using the tool as a skill.

What happens if I use a phrase not listed in the documentation?

Any natural-language request that does not match the six documented trigger phrases will be treated as a generic request. The system will not invoke the specific notebooklm generate commands for audio, PPT, or mind-map creation, instead processing the input through standard NotebookLM analysis workflows.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}