- Use your microphone
- Load 5 models by content hash
App
Voice
Speak to a local model and have it speak back. Pipeline-based voice agent: Whisper STT → Qwen LLM → system speech synthesis. Designed to migrate cleanly to end-to-end omni models when those land.
Capability summary
Derived from the app's manifest. Enforced by the runtime.
- Access the network
- Use your camera
- Record your screen
- Phone home or send telemetry
- Read other files outside its declared scope
Models
Loaded from a content-addressed cache.
qwen2.5-1.5b-instruct-q4_k_m@sha256:6a1a2eb6d15622bf3c96857206351ba97e1af16c30d7a74ee38970e434e9407ewhisper-base.en@sha256:a03779c86df3323075f5e796cb2ce5029f00ec8869eee3fdfb897afe36c6d002moshi-7b@kyutai-moshikopersonaplex-7b@v1-narratorqwen-omni-7b@q4_k_m
Download
v0.0.1 for Apple Silicon Mac.
Signed by Locara CI; provenance attestation available. Universal Apple Silicon binary, ≤ 100 MB before models.
Download .dmg →