OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Audio & Voice AI

Parlor

Parlor is a self-hosted, browser-based application for real-time multimodal voice AI running on a local machine. It combines speech recognition, a Gemma model served through llama.cpp, and Kokoro text-to-speech, with optional camera input, spoken responses, timers, translation and silent transcription modes.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Parlor — OpenViST project profile in Audio & Voice AI.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Parlor when you want to explore a self-hosted, real-time voice interface and have compatible local hardware and dependencies. Configure the optional reasoner only when you want background research through an external endpoint.

Who it is for

  • Developers and self-hosters exploring on-device voice AI.
  • People building or trying local spoken conversation tools, including English-speaking practice.

Documented uses

  • Talk to a local AI using a browser microphone and receive both a transcript and spoken response.
  • Use camera input alongside speech in a multimodal conversation.
  • Ask for a timer, translate spoken utterances, or transcribe speech without spoken replies.
  • Optionally hand a research question to a configured external reasoner while continuing the conversation.
Before you choose

Requirements & limits

The documented application is self-hosted and launched locally at http://localhost:8000. The README describes it as an early research preview.

  • Requires Python 3.12+, llama.cpp build b9503 or newer, and macOS with Apple Silicon or Linux with a supported GPU.
  • The default E4B model requires about 6 GB free RAM; first-run model downloads total about 5.7 GB plus TTS models.
  • The core conversation can remain on-device; background research requires configuring a reasoner API key and OpenAI-compatible endpoint.
  • The repository lists Apache-2.0 licensing and warns that the project may have rough edges and bugs.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Set up Parlor for a local voice session

Install the documented dependencies, launch the server and open its browser interface.

Before you start

  • Python 3.12 or newer
  • llama.cpp build b9503 or newer
  • macOS with Apple Silicon or Linux with a supported GPU
  • About 6 GB free RAM for the default E4B model
  • A browser with camera and microphone access

Steps

  1. Clone https://github.com/fikrikarim/parlor.git and change into the parlor directory.

  2. Install uv and llama.cpp using the project’s documented installation instructions.

  3. Run the command below, then the command below.

    Documented command
    uv sync
    Documented command
    uv run parlor

  4. Open http://localhost:8000, grant camera and microphone access, and start talking.

Expected outcome

The local Parlor browser interface is available for a voice session; models download automatically on first run.

Guide limitations

  • The project labels itself a research preview and warns of rough edges and bugs.
  • The first-run download is about 5.7 GB for the default Gemma model and multimodal projector, plus TTS models.
  • Background research is off unless a reasoner API key is configured.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
14 October 2026
Original website title
fikrikarim/parlor · Official homepage title · retained in its source language
How we review sourcesSuggest a correction