# fikrikarim/parlor

On-device, real-time multimodal AI with features similar to GPT-Live

Repository: https://github.com/fikrikarim/parlor
Canonical: https://ross.abutalabs.com/products/parlor
Language: Python
License: Apache-2.0
License Family: permissive
Topics: gemma, kokoro, litert-lm, local-llm, multimodal, on-device-ai, python, real-time, speech-recognition, text-to-speech, voice-assistant, apple-silicon, mlx
Last push: 2026-08-03T18:06:56+00:00

## Health v2 (maintenance only)
Score: 78/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 95, release rhythm 96, longevity 10
- inputs: {"age_days": 150, "days_push": 30, "days_rel": 31, "gap_med": 4, "n_releases_24m": 2}
- flags: young
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2039, forks 262 (observed 2026-08-28T04:06:08.267996+00:00)

## What it is
Parlor is a fully on-device, real-time multimodal voice assistant similar to GPT-Live, combining speech recognition, a Gemma vision-language model via llama.cpp, and Kokoro text-to-speech in a cascade pipeline. It runs a FastAPI WebSocket server that streams audio and camera frames from a browser, with optional background research via a frontier model API.

## Use cases
- run a private voice assistant entirely on my MacBook
- build a GPT-Live alternative that works offline
- self-host a real-time multimodal AI chatbot
- talk to an AI that can see through my camera and hear me
- hands-free voice conversation with automatic turn-taking
- local voice AI for language speaking practice

## When to choose
- you want a fully local, privacy-preserving voice assistant on Apple Silicon or Linux
- you need real-time multimodal (voice + vision) interaction without cloud APIs
- you want to self-host a GPT-Live-like experience with no per-token costs

## When to avoid
- you need production-grade stability - it is an early research preview with rough edges
- you need a full-duplex speech model rather than a cascade pipeline
- you need Windows or mobile support
- you lack capable local hardware to run a 12B-class model

## Facets
- artifact type: application
- maturity: experimental
- function: speech-recognition, tts, llm-inference, chatbot, http-server, websocket, machine-learning
- domain: artificial-intelligence, speech-processing, large-language-models, chatbots, self-hosted
- platform: python, self-hosted
- tags: voice-assistant, on-device-ai, multimodal, real-time, gemma, kokoro, mlx, apple-silicon, local-llm, smart-turn, macos, linux, web-server

## Member repositories
- fikrikarim/parlor (main) score 78

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:08.267996+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T02:59:12.203849+00:00, confidence not recorded.
  - readme: https://github.com/fikrikarim/parlor (fetched 2026-08-28T04:06:08.267996+00:00, sha 182dea8d31a5)
- Data as of 2026-08-30T08:39:29.467469+00:00.