Natural conversation loop
Press the Converse hotkey, speak, hear the answer, and continue. Silence, a goodbye phrase, or the hotkey ends the session.
Talk to Ollama on Linux
Turn Ollama into a system-wide voice conversation.
Keep the model and the microphone local. AverVOX listens, transcribes, streams the prompt to Ollama’s OpenAI-compatible endpoint, and speaks each response sentence as it arrives.
Ollama /v1 compatibleLocal microphone audioStreaming repliesMultiple model profiles
What it gives you
Voice without wrapping Ollama in another chat application.
AverVOX stays available at the desktop layer, while Ollama remains the model server you already operate.
Sentence-level streaming
AverVOX starts speech when the first complete sentence is available rather than waiting for the model’s entire response.
Profiles for different models
Keep separate labels, endpoints, model selections, and—on Pro—personas for coding, writing, research, or other tasks.
Setup
Connect the model server you already have.
Ollama’s local OpenAI-compatible endpoint normally needs no API key. Pull at least one model before testing the AverVOX profile.
Start Ollama and pull a model
Confirm Ollama is listening locally and that the model you want appears in its model list.
Create an AverVOX profile
Open Settings, add a profile pointing to http://127.0.0.1:11434/v1, leave the local API key blank, test the connection, and select a model.
Press Ctrl+Alt+C
Speak after the listening indicator appears. The reply streams through local TTS; continue naturally or use a goodbye phrase to stop.
No model lock-in
Switch models without rebuilding the voice stack.
The speech layer remains the same while you change an Ollama model, move the endpoint to another machine, or add a second provider.
- Keep several named LLM profiles
- Hot-switch profiles from the tray
- Use local or remote OpenAI-compatible endpoints
- Keep STT and TTS on the Linux desktop
Common questions
What to know before you start.
Does AverVOX replace Ollama?
No. Ollama continues to host and run the model. AverVOX supplies the Linux desktop speech, conversation loop, streaming TTS, profiles, and optional agent-facing bridge.
Does any conversation leave my network?
Not when the configured Ollama model is local. Speech processing is local, and conversation text is sent only to the local endpoint you enter. Ollama cloud models or remote endpoints follow their own network behavior.
Can Ollama run on another computer?
Yes. Point the profile at the reachable LAN address of an OpenAI-compatible Ollama endpoint and secure that server appropriately. AverVOX speech can remain local to the desktop machine.
What happens if a model stalls?
AverVOX monitors first-token, empty-response, and stream-read timeouts. Its documented defaults are designed to stop a broken voice loop rather than leave you waiting indefinitely.
Can I interrupt a long answer?
Yes. Voice interruption is available, with headphones recommended to prevent the assistant’s own speech from feeding back into the microphone.
Start with the workflow
Keep Ollama local. Make the conversation audible.
Install AverVOX OSS for the complete core voice loop, then add Pro when you want Kokoro, wake-word activation, persistent sessions, personas, and LAN voice sharing.