her-voice
by matusvojtek
Speech & TranscriptionBrief
Give your agent a voice.
Description
What This Skill Does
Her Voice adds spoken audio responses to an agent using Kokoro TTS, a local 82M-parameter neural model. Audio streams on-device with no API keys or cloud dependencies. Supports a real-time LED visualizer, a background daemon for low latency, and voice blending between multiple voices.
Runs entirely on-device with no API keys or cloud costs, while still achieving sub-second first-audio latency via streaming and an optional warm-model daemon.
When to Use It
- Hearing agent replies read aloud while working hands-free
- Listening to long code summaries without reading the screen
- Using persist mode as a standalone voice station for pasted text
- Saving spoken responses as WAV files for later playback
- Getting instant audio feedback when the agent starts responding
Example Workflow
User asks: Read this summary aloud
- 1
Check if voice mode is enabled or requested by the user
- 2
Play a macOS notification sound immediately for instant feedback while TTS generates
- 3
Call speak.py with the response text
- 4
Stream audio output while the visualizer animates in real-time
- 5
Display the text response alongside the audio for accessibility
Spoken audio plays through speakers with a floating LED visualizer overlay; text response also shown in the terminal
Requirements
- espeak-ng installed (brew install espeak-ng on macOS, apt install espeak-ng on Linux)
Installation
clawhub install matusvojtek/her-voiceRun the CLI command, or paste the prompt into Claude Code, Cursor, Codex or whichever coding agent you use.