MetaHuman twin
Photorealistic avatar in Unreal Engine 5. Your face or a custom character.
MetaHuman · Unreal Engine 5 · RAG · Ollama
Digital twins that speak, listen and remember.
We build photorealistic MetaHuman avatars connected to an AI brain (RAG + memory + LLM). Sports coach, influencer, technician: same engine, different personality.
click Click any block to hear it
Services
AI brain · RAG + memory
Your documents become knowledge. Ollama locally or API. Per-visitor memory.
Real-time Pixel Streaming
WebRTC. The visitor sees a live Unreal render in the browser, no install.
n8n orchestration
CRM, WhatsApp, email, booking. The twin acts, it does not just talk.
Unreal Engine plugins
Three plugins for Unreal Engine 5. Each one drops into your own project, drives your own MetaHuman, and runs on your machine — no subscription, no account, no per-minute fees.
SinaLink — Talking MetaHuman
Talk to your MetaHuman and she answers out loud. Real-time lip-sync driven by the words themselves, gestures that fit the sentence, eye contact down the lens. For receptions, kiosks, websites and in-game characters.
In review on Fab — ask for early accessSinaLink Presenter — Text to Talking MetaHuman
Paste a script, she performs it. Inline tags control pauses, mood and where she looks. Batch mode queues twenty scripts and arms Take Recorder around every take, so a folder of finished performances is waiting in the morning.
In review on Fab — ask for early accessSinaLink Live — Stream Co-Host
She reads your Twitch and YouTube chat and answers out loud, live. Donations are thanked by name first, nobody gets two turns in a row, and stale messages are dropped rather than queued. She reads chat — she never types in it.
In review on Fab — ask for early accessProcess
-
1 · Scan & modeling
Photos or 3D scan → MetaHuman Creator → ARKit facial rig.
-
2 · Knowledge ingestion
PDF, site, FAQ, videos → chunks → embeddings → Qdrant / pgvector.
-
3 · Brain & voice
Ollama / API → streamed text → sentence-by-sentence TTS → lip-sync.
-
4 · Delivery
Web (Pixel Streaming), kiosk, WhatsApp, or pre-rendered video.
Why Sina
Your data stays with you
Ollama runs on your machine. No document leaves your network.
Under 1.5 s
From click to lip movement. Past 3 s, the visitor leaves.
Darija · French · English
The twin switches language mid-conversation without losing memory.
Articles
What I learned building this, written down properly — including the parts that went wrong.
What a digital twin actually is (and when you do not need one)
The difference between a video avatar, a deepfake and a digital twin — and the one question to answer before anyone builds you one.
The real cost of an AI avatar, per minute and per year
Hosted platforms charge $0.10 to $0.37 a minute. Here is what that becomes over a trade show, a stream schedule and a permanent installation.
How MetaHuman lip-sync actually works
Why following the volume of the audio gives you a flapping jaw, and why the mouth has to move sixty milliseconds before the sound.
Let's talk about your twin
Describe your use case. We reply within 24 h with a demo of your persona.
Or email t81251675@gmail.com — one sentence about what it is for is enough to start.