Matilda Ears

Speech-to-text engine with multiple operation modes, real-time transcription, and Ears Tuner formatting.

Quick Start

pipx install .
./scripts/setup.sh install --dev
ears status

Basic Usage

ears status
ears models
ears status --json
./scripts/test.py --help

Operation Modes

• Listen-once
• Conversation
• Tap-to-talk / Hold-to-talk
• File transcription
• WebSocket server
• Wake word

Mode implementations live in src/matilda_ears/modes/.

Wake Word

Wake word models live in src/matilda_ears/wake_word/models/.

ears --wake-word --agent-aliases="Matilda:hey_jarvis"
ears train-wake-word "hey matilda"

Configuration

Configuration lives in ~/.matilda/config.toml under the [ears] section.

Ears Tuner settings live under [ears.ears_tuner]:

[ears.ears_tuner]
enabled = true
formatter = "pipeline"

[ears.ears_tuner.filename_formats]
md = "UPPER_SNAKE"
json = "lower_snake"
py = "lower_snake"
js = "camelCase"
"*" = "lower_snake"

ears status --json
ears models --json

Development

./scripts/setup.sh install --dev
python -m spacy download en_core_web_sm

make test
make quality

Documentation

• docs/
• AGENTS.md
• CHANGELOG.md

Related Projects

• Matilda
• Matilda Voice
• Matilda Brain

License

MIT License