speak

Text-to-speech CLI (speak): Groq first when it supports the language, then local Orpheus, then macOS say.

English:  text → Groq Orpheus → local Orpheus → macOS say
Arabic:   text → Groq Orpheus → macOS say
German:   text → local Orpheus (DE) → macOS say
Other:    text → macOS say

Forced --engine groq|local|say never falls back.

Quick start

uv tool install git+https://github.com/thaikolja/speaker-cli.git

mkdir -p ~/.config/speak
speak --write-config ~/.config/speak/config.json
# set groq_api_key (or export GROQ_API_KEY)

speak "Hello from Groq."
speak "Guten Tag."
speak -d cheerful "Great news!"
speak -f notes.txt --speed 1.25
speak --engine say "Offline only"
speak --help

Documentation

Doc	Contents
docs/architecture.md	Modules, chain, data flow
docs/configuration.md	config.json, env, settings
docs/backends.md	Groq / local / say
docs/cli-reference.md	Flags, exit codes, examples
docs/development.md	Setup, tests, gates
docs/troubleshooting.md	Common failures

Config

JSON only (not .env for defaults). Search order: --config → $SPEAK_CONFIG → ./config.json → ~/.config/speak/config.json.

Optional env: GROQ_API_KEY, SPEAK_CONFIG. See .env.example and config.example.json.

Models

Path	Language	Model/Asset	Default Voice
Groq	English	canopylabs/orpheus-v1-english	troy (max 200 chars)
Groq	Arabic	canopylabs/orpheus-arabic-saudi	fahad
Local	English	GGUF EN	leo
Local	German	GGUF DE	leo
say	system	macOS voices	auto

Offline/local Orpheus (optional)

git clone https://github.com/thaikolja/speaker-cli.git
cd speaker-cli
uv sync --extra dev
make install-metal   # Apple Silicon Metal llama-cpp-python
speak --engine local "Hello offline."

Development

uv sync --extra dev
make check

See docs/development.md.

License

MIT