OpenAI-compatible connection
Point AverVOX at LM Studio’s local /v1 base URL, test the endpoint, and choose one of the models the server exposes.
Voice for LM Studio on Linux
Talk to LM Studio from anywhere on your desktop.
Use LM Studio for local model inference and AverVOX for system-wide microphone input, streaming speech output, hotkeys, profiles, and hands-free workflows.
LM Studio /v1 endpointLocal STT and TTSModel auto-discoveryX11 / XWayland
What it gives you
Keep LM Studio focused on models. Let AverVOX own the voice workflow.
The two products meet at a standard local API, so each can do the job it is best at.
Streaming local speech
Replies are cleaned of distracting markup and sent to TTS sentence by sentence for faster time to first audio.
Available beyond the LM Studio window
Dictation and selected-text TTS remain system-wide, so one installation also helps in editors, browsers, terminals, and email.
Setup
Start the server, select a model, speak.
LM Studio’s Developer area can expose local models on the default port, and its CLI can start the server as well.
Start the LM Studio API server
Load or make a model available, then start the local server in the Developer tab or with the LM Studio CLI.
Add the endpoint in AverVOX
Use http://localhost:1234/v1, enter a token only if authentication is enabled, test the connection, and select a visible model.
Begin Converse
Press Ctrl+Alt+C. AverVOX handles recording, VAD, local STT, streaming model output, and local TTS around the LM Studio endpoint.
Desktop-wide utility
One connection does more than voice chat.
Even when LM Studio supplies the model, AverVOX remains useful independently of it.
Dictate prompts where you write them
Speak into a browser, source editor, issue tracker, email client, or terminal. The LLM does not need to be involved in simple dictation.
Listen to selected text
Use local TTS to review a draft or read documentation aloud.
Hot-switch models
Change the active model profile from the tray without rebuilding your desktop workflow.
Separate the speech machine from the inference machine
AverVOX can run speech on an ordinary Linux laptop while LM Studio serves a larger model from another machine on the LAN.
Common questions
What to know before you start.
Does LM Studio have to stay in the foreground?
No. AverVOX communicates with its API server. LM Studio can serve locally while you work in other applications; headless server options depend on your LM Studio setup.
Which endpoint should I enter?
Use the OpenAI-compatible base URL, normally http://localhost:1234/v1. This is distinct from LM Studio’s native REST API paths.
Can AverVOX use a model on another LAN machine?
Yes, when the LM Studio server is deliberately exposed on the network and protected appropriately. Enter that reachable base URL instead of localhost.
Can I keep several LM Studio model profiles?
Yes. AverVOX supports multiple profiles and model selections. Pro adds a separate system prompt/persona and persistent context for each profile.
Will it speak reasoning markup and Markdown syntax?
AverVOX strips common Markdown from spoken output so the reply sounds more natural rather than reading punctuation and formatting tokens aloud.
Start with the workflow
Make your LM Studio server conversational across Linux.
The open-source edition supplies the full local voice loop. Pro adds a more natural Kokoro voice, wake-word activation, profile personas, persistent sessions, and LAN sharing.