TTS Setup and Configuration

TTS Setup and Configuration for Claude Code

Provide comprehensive guidance for setting up, configuring, and troubleshooting the TTS (text-to-speech) plugin for Claude Code. This skill covers installation prerequisites, configuration options, voice selection, troubleshooting, and usage patterns.

Prerequisites and Installation

kokoro-tts Installation

Before using this plugin, kokoro-tts must be installed. Guide the user through verification:

1. Check if kokoro-tts is installed:

   which kokoro-tts
   

2. If not installed, direct user to installation:

   • Official repository: https://github.com/thewh1teagle/kokoro
   • Follow installation instructions for their operating system
   • Verify installation: kokoro-tts --help

3. Verify model files exist:

   • Model: ~/.local/share/kokoro-tts/kokoro-v1.0.onnx
   • Voices: ~/.local/share/kokoro-tts/voices-v1.0.bin
   • If missing, kokoro-tts installation may be incomplete

Plugin Installation

The TTS plugin should already be installed if the user is asking about it. If not:

1. Install plugin to Claude Code plugins directory
2. Enable in Claude Code settings or with cc --plugin-dir
3. Restart Claude Code for hooks to load

Initial Configuration

Quick Start

Guide user through quick configuration:

1. Run the configure command:

   /tts-plugin:configure
   

2. Choose "Quick Setup" for common settings:

   • TTS_ENABLED: true (to enable TTS)
   • TTS_VOICE: Select preferred voice (af_bella, af_sarah, etc.)
   • TTS_LANG: Select language (en-gb, en-us, etc.)
   • TTS_SPEED: Set speech speed (1.0-1.5 recommended)

3. Save and restart Claude Code for changes to take effect

Test TTS

After configuration, test that it works:

/tts-plugin:test

Expected: Should hear "This is a test of the text to speech system"

Voice and Language Selection

Available Voices

Run kokoro-tts --help-voices to see all available voices.

Common voices:

• af_bella - Female, British accent (default)
• af_sarah - Female, British accent
• af_sky - Female, British accent
• bf_emma - Female, British accent
• bf_isabella - Female, British accent
• am_adam - Male, British accent
• am_michael - Male, British accent

Available Languages

Run kokoro-tts --help-languages to see all available languages.

Common languages:

• en-gb - English (British) (default)
• en-us - English (American)
• es - Spanish
• fr - French
• de - German
• ja - Japanese

Testing Different Voices

Test voices without changing configuration:

/tts-plugin:test --voice af_sarah
/tts-plugin:test --voice bf_emma --speed 1.5
/tts-plugin:test "Custom message" --voice am_michael

Once you find a voice you like, save it with /tts-plugin:configure.

Configuration Options

Basic Settings

TTS_ENABLED (true/false)

• Master enable/disable switch
• Affects all TTS hooks
• Default: true

TTS_VOICE (voice name)

• Voice to use for speech
• See kokoro-tts --help-voices for options
• Default: af_bella

TTS_LANG (language code)

• Language for speech
• See kokoro-tts --help-languages for options
• Default: en-gb

TTS_SPEED (0.5-2.0)

• Speech speed multiplier
• 1.0 = normal speed
• 1.3-1.5 = faster but still clear (recommended)
• Default: 1.3

Advanced Settings

TTS_PRETOOL_ENABLED (true/false)

• Enable PreToolUse hook (speaks before each tool execution)
• Can be verbose, disabled by default
• Only active if TTS_ENABLED=true
• Default: false

TTS_USE_TTS_SECTION (true/false)

• Extract "## TTS Response" section from Claude's responses
• When true, only speaks TTS-optimized content
• When false, speaks entire response
• Default: true

TTS_MAX_LENGTH (number)

• Maximum characters to speak per message
• Prevents extremely long TTS
• Default: 5000

TTS_STATE_DIR (path)

• Directory for session state files
• Tracks what's been spoken to avoid repeats
• Default: ~/.local/state/claude-tts/session-state

TTS_LOG_DIR (path)

• Directory for TTS log files
• Useful for debugging
• Default: ~/.local/state/claude-tts/logs

Configuration File Location

All settings stored in: ~/.claude/tts-plugin.env

Edit this file directly or use /tts-plugin:configure for interactive setup.

How TTS Works

Event Flow

1. User submits prompt (UserPromptSubmit hook):

   • Interrupts any ongoing TTS playback
   • Injects instruction asking Claude to add "## TTS Response" section
   • Marks current messages as read

2. Claude generates response:

   • Includes TTS-optimized content in "## TTS Response" section
   • This section has markdown removed, simplified paths, natural language

3. Claude finishes responding (Stop hook):

   • Extracts new messages from transcript
   • Looks for "## TTS Response" section
   • Speaks the text using kokoro-tts
   • Updates session state

4. Optional: Before each tool (PreToolUse hook, if enabled):

   • Speaks any new text that appeared
   • Provides real-time audio feedback

5. Session ends (SessionEnd hook):

   • Cleans up session state files

TTS Response Section

Claude adds a "## TTS Response" section to responses when TTS is active. This section contains:

• Same information as main response
• Markdown formatting removed
• URLs converted to domain names
• File paths simplified to filenames
• Natural, spoken language

Users see both the normal response AND the TTS Response section. TTS only speaks the TTS Response section (if TTS_USE_TTS_SECTION=true).

Common Use Cases

Enable TTS for Session

Temporarily enable TTS without saving:

/tts-plugin:enable

Enable TTS Permanently

Save enable setting to .env:

/tts-plugin:enable --persistent

Disable TTS Temporarily

Stop TTS for current session:

/tts-plugin:disable

Disable TTS Permanently

Save disable setting to .env:

/tts-plugin:disable --persistent

Change Voice

/tts-plugin:configure
# Choose Quick Setup
# Select new voice

Adjust Speed

/tts-plugin:configure
# Choose Quick Setup
# Set new speed (1.0-2.0)

Enable PreToolUse Hook

For real-time TTS during tool execution:

/tts-plugin:configure
# Choose Advanced Setup
# Set TTS_PRETOOL_ENABLED=true

Note: This can be verbose as it speaks before each tool.

Troubleshooting

TTS Not Speaking

Problem: TTS is enabled but nothing plays

Solutions:

1. Verify kokoro-tts is installed: which kokoro-tts
2. Test TTS directly: /tts-plugin:test
3. Check TTS_ENABLED in configuration: /tts-plugin:configure
4. Verify system audio is working
5. Check logs: ~/.local/state/claude-tts/logs/
6. Restart Claude Code (hooks load at startup)

TTS Interrupted Immediately

Problem: TTS starts but stops right away

Solutions:

1. Don't submit new prompts while TTS is speaking
2. UserPromptSubmit hook intentionally interrupts TTS
3. This is normal behavior to prevent overlapping speech

Wrong Voice or Language

Problem: TTS uses different voice than expected

Solutions:

1. Check current configuration: /tts-plugin:configure → View Current Settings
2. Update configuration: /tts-plugin:configure → Quick Setup
3. Restart Claude Code after changes
4. Test new settings: /tts-plugin:test

TTS Too Fast/Slow

Problem: Speech speed is uncomfortable

Solutions:

1. Test different speeds: /tts-plugin:test --speed 1.0 (try 0.8-1.5)
2. Save preferred speed: /tts-plugin:configure → Quick Setup → Set TTS_SPEED
3. Restart Claude Code after changes

TTS Too Verbose

Problem: TTS speaks too much or too often

Solutions:

1. Disable PreToolUse hook: /tts-plugin:configure → Advanced → TTS_PRETOOL_ENABLED=false
2. Reduce max length: /tts-plugin:configure → Advanced → TTS_MAX_LENGTH=2000
3. Disable completely: /tts-plugin:disable --persistent

TTS Speaks Full Response

Problem: TTS speaks entire response instead of optimized section

Solutions:

1. Ensure TTS_USE_TTS_SECTION=true in configuration
2. Verify Claude is adding "## TTS Response" section (check responses)
3. If section missing, TTS instruction injection may not be working
4. Check UserPromptSubmit hook is active (restart Claude Code)

Model Files Not Found

Problem: Error about missing model or voices files

Solutions:

1. Verify model file exists: ls ~/.local/share/kokoro-tts/kokoro-v1.0.onnx
2. Verify voices file exists: ls ~/.local/share/kokoro-tts/voices-v1.0.bin
3. If missing, reinstall kokoro-tts
4. Update paths in configuration if files are in different location

Changes Not Taking Effect

Problem: Configuration changes don't work

Solutions:

1. Restart Claude Code - hooks and configuration load at startup
2. Verify changes saved to ~/.claude/tts-plugin.env
3. Check for syntax errors in .env file
4. Test with /tts-plugin:test to verify settings

Best Practices

Recommended Settings

For most users:

TTS_ENABLED=true
TTS_PRETOOL_ENABLED=false       # Disable verbose PreToolUse
TTS_VOICE=af_bella              # Or your preferred voice
TTS_LANG=en-gb                  # Or your preferred language
TTS_SPEED=1.3                   # Faster but clear
TTS_USE_TTS_SECTION=true        # Use optimized content
TTS_MAX_LENGTH=5000             # Prevent overly long TTS

When to Enable PreToolUse Hook

Enable if you want:

• Real-time audio feedback during long operations
• To hear progress updates as tools execute
• Maximum awareness of what Claude is doing

Disable if you want:

• Less verbose TTS (recommended)
• Only hear final responses
• Avoid hearing tool execution messages

Speed Recommendations

• 0.8-1.0: Slower, easier to understand, good for learning
• 1.0-1.3: Normal speed, comfortable listening
• 1.3-1.5: Faster but clear, efficient (recommended)
• 1.5-2.0: Very fast, may be hard to follow

Voice Selection Tips

1. Test multiple voices: /tts-plugin:test --voice VOICE_NAME
2. Consider accent preference (British vs American)
3. Consider gender preference (male vs female voices)
4. Listen for clarity at your preferred speed
5. Save favorite: /tts-plugin:configure

Advanced Topics

Temporarily Disabling TTS

To quickly silence TTS without changing configuration:

/tts-plugin:disable

Re-enable when ready:

/tts-plugin:enable

Session State Management

TTS tracks what's been spoken in session state files:

• Location: ~/.local/state/claude-tts/session-state/
• Files named: claude-tts-state-{session_id}.txt
• Automatically cleaned up on SessionEnd
• Prevents re-speaking already-heard messages

Log Files

TTS logs to:

• Location: ~/.local/state/claude-tts/logs/
• Files named: claude-tts-{session_id}.log
• Useful for debugging
• Check these if TTS isn't working

Custom Model/Voices Paths

If kokoro-tts files are in non-standard location:

/tts-plugin:configure
# Choose Advanced Setup
# Set TTS_MODEL=/custom/path/to/model.onnx
# Set TTS_VOICES=/custom/path/to/voices.bin

Quick Reference

Commands

• /tts-plugin:enable [--persistent] - Enable TTS
• /tts-plugin:disable [--persistent] - Disable TTS
• /tts-plugin:configure - Configure settings
• /tts-plugin:test [message] [--voice V] [--speed S] [--lang L] - Test TTS

Configuration File

• Location: ~/.claude/tts-plugin.env
• Template: Plugin includes .env.example
• Edit: Use /tts-plugin:configure or edit file directly

Key Settings

• TTS_ENABLED: Master on/off switch
• TTS_VOICE: Voice selection
• TTS_SPEED: Speech speed (0.5-2.0)
• TTS_PRETOOL_ENABLED: Enable verbose PreToolUse hook

Getting Help

1. Ask about TTS setup: "How do I configure TTS?"
2. Run test: /tts-plugin:test
3. Check configuration: /tts-plugin:configure → View Current Settings
4. Check logs: ~/.local/state/claude-tts/logs/
5. Check kokoro-tts: kokoro-tts --help

---

When helping users with TTS:

1. Start by understanding their issue or goal
2. Guide them through relevant commands
3. Provide specific troubleshooting steps
4. Test changes with /tts-plugin:test
5. Remind to restart Claude Code after configuration changes