Free Tool Lets You Talk to AI Like a Person, No Cloud Needed
This could make voice assistants faster and more private—no internet required.
VoiceManager is a Trek-inspired, locally running speech-to-speech app with low-latency streaming over one WebSocket. The developer started it in March to support a game concept and AI assistants.
Its default configuration uses a special audio.cpp sidecar with Nemotron and Breeze TTS 2. It supports turn-based conversations with barge-in, which is off by default because echo cancellation is tricky; after a tablet update broke hardware AEC, the developer added a toggle.
VoiceManager includes client integration docs and a Python client, and the Kotlin console app serves as a client reference. In the video, a wake word (“hey computer”) starts a conversation and runs on the tablet, replies stream back as speech sentence by sentence, images and markdown documents show on the panel under the reply, tool calls go back to the app and display volume
- VoiceManager lets you talk to AI on your own device, so your voice data stays private and responses are fast.
- You can interrupt the AI mid-sentence, and it remembers past conversations for a more natural feel.
- It's free and open source, but setting it up requires some technical skills and a capable device.
Why It Matters
This could lead to voice assistants that respect your privacy and work without an internet connection.