Piper TTS — Local Voice Messages
Generate voice messages using Piper, a fast local TTS engine. Zero cloud calls, zero cost, zero API keys.
Setup
If Piper is not installed, run the setup script:
scripts/setup-piper.sh
This installs piper-tts via pip and downloads a default voice (en_US-kusal-medium).
Generating Voice Messages
Use scripts/piper-speak.sh to generate and deliver voice:
scripts/piper-speak.sh "<text>" [voice]
text: The text to speak (required)voice: Piper voice name (default:en_US-kusal-medium)
The script outputs an MP3 path. Include it in your reply as:
[[audio_as_voice]]
MEDIA:<path-to-mp3>
This delivers the audio as a native voice message on supported channels (Telegram, Discord, etc.).
Example Workflow
- User asks: "Tell me a joke as audio"
- Run:
scripts/piper-speak.sh "Why do programmers prefer dark mode? Because light attracts bugs!" - Get MP3 path from output
- Reply with
[[audio_as_voice]]+MEDIA:<path>
Available Voices
After setup, download additional voices:
scripts/setup-piper.sh --voice en_US-ryan-high
scripts/setup-piper.sh --voice en_GB-northern_english_male-medium
Popular voices:
en_US-kusal-medium— Clear male voice (default, recommended)en_US-ryan-high— High quality US maleen_US-hfc_male-medium— US maleen_GB-northern_english_male-medium— British male- Browse all: https://huggingface.co/rhasspy/piper-voices
Important Notes
- Speed: Local generation is ~0.5-1s. Much faster than cloud TTS.
- No API keys: Works completely offline after setup.
- Platform: macOS (Apple Silicon + Intel), Linux. Requires Python 3.9+.
- Do NOT set
messages.tts.auto: "always"in OpenClaw config — it makes every response slow. Keep TTS on-demand.
