The open-source video meeting platform you can self-host in 60 seconds. Real-time AI captions in 100+ languages — running entirely in your browser.
docker compose up -dOpen http://localhost:3000. That is it.
Live demo: meet.liqaa.io. Full self-hosting guide (TLS, Caddy, nginx, sizing) in SELF_HOSTING.md.
Zoom is bloated. Google Meet locks you into Google. Jitsi is functional but unloved. Microsoft Teams ships malware-grade tracking. Whereby costs $90/month for a small team.
In 2026, the world's video infrastructure deserves to be:
- Open source — read every line, fork it, ship your own
- Self-hostable — your data, your servers, your rules
- Beautiful — Stripe-grade design isn't optional anymore
- AI-native — captions, summaries, translation in 100+ languages
- Privacy-first — captions run in your browser, never sent to a server
- Fast — Next.js 16 + LiveKit SFU + WASM = sub-second time-to-video
That's LIQAA Meet.
- No-install join — share a link, click, you're in
- HD video, simulcast — adapts to bad connections
- Screen share with audio (Chromium + Edge + Firefox)
- Raise hand, react with emojis, in-call chat
- Breakout rooms for up to 50 simultaneous splits
- Cloud or self-host — same experience, your choice
| LIQAA Meet | Zoom | Google Meet | Jitsi | |
|---|---|---|---|---|
| In-browser AI captions (no server) | Whisper.cpp WASM + WebGPU | (server-side, paid) | (server-side) | |
| AI summary after the call | local Ollama or any API | ($14/mo Pro) | (Workspace add-on) | |
| Self-hostable | docker compose up |
(complex) | ||
| End-to-end encrypted by default | (only paid Pro+) | partial | optional | |
| Bring your own LLM | Ollama, OpenAI, Anthropic, llama.cpp | |||
| Open source | AGPL-3.0 | Apache-2.0 | ||
| Cost (10 users, 100 hrs/mo) | $0 self-hosted | $150 | $72 | $0 |
Open https://meet.liqaa.io
Click "New room"
Share the link
That's it. No account. No download.
You need: Docker. That's all.
git clone --depth 1 https://github.com/hartemyaakoub/liqaa-meet.git
cd liqaa-meet
cp .env.example .env
# Edit .env: set MEET_DOMAIN to your domain (or "localhost:3000")
docker compose up -dOpen http://localhost:3000. Done.
For production deployment with TLS, see SELF_HOSTING.md.
Embed LIQAA Meet rooms in your own SaaS:
import { MeetRoom } from '@liqaa/meet';
<MeetRoom
roomCode="abc-defg-hij"
user={{ name: 'Alice', email: 'alice@example.com' }}
features={{ captions: true, summary: true, recording: false }}
liqaaEndpoint="https://your-self-hosted-liqaa.com"
/>Most video platforms send your audio to their servers, transcribe it there, send it back. We don't.
Your microphone
Web Audio API → 16 kHz PCM
Web Worker (off main thread)
Whisper.cpp compiled to WebAssembly
WebGPU acceleration (fallback: WASM SIMD)
Captions render locally
The Whisper model (39 MB for tiny, 142 MB for base) is cached after first load. Inference: ~200 ms for 1 second of audio on M1 / Ryzen 5 laptops. Your audio never leaves the browser.
Languages: Whisper supports 99 — Arabic, French, English, Spanish, Mandarin, German, Russian, Hindi, Portuguese, Japanese, Korean, Italian, Turkish, Polish, Dutch, Hebrew, Greek, Persian, Swahili — and 80 more.
After the call, the captions transcript is sent to your configured LLM:
- Default: local Ollama (
llama3.2:3b) — runs on your laptop, free, private - Or: OpenAI / Anthropic / Google / any OpenAI-compatible API — set
LLM_PROVIDER=openai - Or: nothing — turn it off entirely with
FEATURE_SUMMARY=false
The summary covers:
- Decisions made — bulleted list
- Action items — assigned to the person who said them
- Open questions — flagged for follow-up
- Full transcript — searchable, editable, exportable as .md / .pdf
meet.liqaa.io
(this repo, Next.js)
/api/sdk-token
LIQAA Cloud
(or self-host)
JWT
LiveKit SFU
(eu-central +
dz-algiers)
/api/rooms SQLite
(via Drizzle)
/api/summary Ollama / OpenAI / Anthropic
For the full design, see docs/architecture.md.
Three deployment recipes:
- Single VPS — one
docker compose up, suitable for ≤500 concurrent users - Kubernetes — Helm chart at
deploy/helm/(community-maintained) - Vercel + LIQAA Cloud — push to Vercel, point at
api.liqaa.io, done
Full guide: SELF_HOSTING.md.
- v0.1 — basic rooms, screen share, chat
- v0.2 — in-browser whisper captions
- v0.3 — AI summary
- v0.4 — recording with auto-uploaded MP4 + transcript
- v0.5 — virtual backgrounds (MediaPipe in browser)
- v0.6 — live translation (Whisper → 99 languages → display)
- v0.7 — calendar invites, ICS export, scheduling
- v1.0 — production-ready release with SLA-eligible self-host package
Track issues at github.com/hartemyaakoub/liqaa-meet/issues.
LIQAA Meet uses the LIQAA video API for media routing. The meet.liqaa.io hosted version runs on LIQAA Cloud; self-hosted versions can either:
- Use LIQAA Cloud (free up to 1,000 minutes/month) — set
LIQAA_PKandLIQAA_SKin.env - Use your own LIQAA self-host — clone hartemyaakoub/liqaa-architecture for the full stack
You can also fork LIQAA Meet and swap LIQAA for any LiveKit-compatible SFU.
We chose AGPL-3.0 (not MIT, not Apache) on purpose:
- You can self-host LIQAA Meet for your company — for free, forever.
- You can fork it and modify it — go wild.
- You can build a paid SaaS on top of it — but you must publish your modifications.
The AGPL is the same license as Cal.com, Plausible Analytics, Mastodon. It guarantees the project stays open. If you need a commercial-friendly license for a closed product, contact us.
We welcome PRs. Read CONTRIBUTING.md — TL;DR: open an issue first for non-trivial work, follow the linter, ship tests when you ship code.
If LIQAA Meet helps you, the cheapest way to support us is a . It actually moves the needle in GitHub trending.
AGPL-3.0 — © 2026 TKAWEN, Algeria. Built in for the world.