Text-to-Speech Skill

Convert text to speech, supporting podcast script parsing, emotion tag processing, and voice-changer post-processing.

Usage Instructions

When users request to convert text to speech, use the following commands:

bash

# Basic usage
python3 ~/.claude/skills/text-to-speech/scripts/text_to_speech.py <text-file>

# Specify output file
python3 ~/.claude/skills/text-to-speech/scripts/text_to_speech.py script.txt -o output.mp3

# Specify voice
python3 ~/.claude/skills/text-to-speech/scripts/text_to_speech.py script.txt -v zh-CN-XiaoxiaoNeural

# Enable post-processing (voice-changer)
python3 ~/.claude/skills/text-to-speech/scripts/text_to_speech.py script.txt --post-process

# List all available voices
python3 ~/.claude/skills/text-to-speech/scripts/text_to_speech.py --list-voices

Core Features

1. Script Parsing

Automatically identify and remove comments and tags in podcast scripts:

Timestamps:
```
(00:00)
```
BGM comments:
```
[BGM渐入：...]
```
(keep original tag format)
Stage directions:
```
(主播声音：...)
```
```
(停顿 1秒)
```
(keep original tag format)
Emotion tags:
```
(语速放慢，加重语气)
```
(keep original tag format)
Markdown tags:
```
**文本**
```

2. Multi-Voice Support

Supports over 18 Chinese voices, including:

Male Voices: YunyangNeural (news broadcasting), YunxiNeural (young and energetic), YunjianNeural (mature and steady)
Female Voices: XiaoxiaoNeural (gentle and friendly), XiaoyiNeural (lively and cheerful), XiaoyanNeural (news broadcasting)

3. Speech Parameter Adjustment

Speech rate adjustment:
```
--rate "+20%"
```
or
```
--rate "-10%"
```
Pitch adjustment:
```
--pitch "+5Hz"
```
or
```
--pitch "-3Hz"
```
Volume adjustment:
```
--volume "+20%"
```
or
```
--volume "-10%"
```

4. Post-Processing Integration

Optional integration with voice-changer skill for voice transformation.

Configuration File

The configuration file is located at:

~/.claude/skills/text-to-speech/config/tts_config.json

Main Configuration Items

json

{
  "edge_tts": {
    "voice": "zh-CN-YunyangNeural",
    "rate": "+0%",
    "pitch": "+0Hz",
    "volume": "+0%"
  },
  "script_parsing": {
    "enabled": true,
    "remove_timestamps": true,
    "remove_bgm_notes": true,
    "remove_stage_directions": true,
    "remove_markdown": true
  },
  "emotion_processing": {
    "enabled": true,
    "use_ssml": true
  },
  "output": {
    "format": "mp3",
    "default_output_dir": "same_as_input",
    "filename_suffix": "_tts"
  },
  "post_processing": {
    "enabled": false,
    "voice_changer": {
      "enabled": false,
      "voice_type": "female_1",
      "pitch_shift": 0
    }
  }
}

Workflow

Input Text/File
    ↓
Script Parsing (remove comments and tags)
    ↓
Emotion Tag Processing (optional)
    ↓
Edge TTS Speech Synthesis
    ↓
Post-Processing (voice-changer, optional)
    ↓
Output MP3 File

Dependency Installation

bash

# Install Edge TTS
pip install edge-tts

# If post-processing is needed, ensure voice-changer skill is installed

Usage Examples

Example 1: Convert Podcast Script

bash

python3 ~/.claude/skills/text-to-speech/scripts/text_to_speech.py podcast_script.txt

The script will automatically:

Remove timestamps and BGM comments
Remove stage directions
Keep only the actual text to be read
Generate
```
podcast_script_tts.mp3
```

Example 2: Use Female Voice and Adjust Speech Rate

bash

python3 ~/.claude/skills/text-to-speech/scripts/text_to_speech.py script.txt \
  -v zh-CN-XiaoxiaoNeural \
  --rate "+10%"

Example 3: Enable Post-Processing

bash

python3 ~/.claude/skills/text-to-speech/scripts/text_to_speech.py script.txt \
  --post-process

This will first generate the speech, then call voice-changer for voice transformation.

Notes

Network Requirement: Edge TTS requires an internet connection
Text Length: It is recommended to convert no more than 10,000 words at a time
Script Format: Supports plain text and annotated podcast scripts
Post-Processing: Requires voice-changer skill to be installed first

Technical Implementation

TTS Engine: Microsoft Edge TTS (free, high-quality)
Script Parsing: Regular expression matching
Audio Format: MP3 (default)
Post-Processing: Optional integration with voice-changer

Performance Reference

1000 words: Approximately 10-20 seconds
5000 words: Approximately 30-60 seconds
Network speed has a significant impact

License

MIT

text-to-speech

NPX Install

Tags

SKILL.md Content (Chinese)

Text-to-Speech Skill

Usage Instructions

Core Features

1. Script Parsing

2. Multi-Voice Support

3. Speech Parameter Adjustment

4. Post-Processing Integration

Configuration File

Main Configuration Items

Workflow

Dependency Installation

Usage Examples

Example 1: Convert Podcast Script

Example 2: Use Female Voice and Adjust Speech Rate

Example 3: Enable Post-Processing

Notes

Technical Implementation

Performance Reference

License