Thank you, and goodbye.
Hi,
Anima Machines has shut down. The servers are off, the database is gone, and the sign-up button is no more. This page is what remains: a short note on what it was, and a gallery of what people made with it.
What it was. Anima Machines let you turn a photo of yourself into an expressive, rigged 3D avatar. The avatar could show emotion in real time, speak, and be driven by a language model, so it could stand in for you, or for an AI, with a face.
Where it came from. It started as the avatar side of emodrink, our paper for Augmented Humans 2026. At the conference we ran a live booth: visitors answered a few questions about their sleep and mood, got a lookalike avatar generated on the spot, and watched it recommend them a drink in English or Japanese. Afterwards the same setup went to the NUS Open House, and then it became a thing we mostly played with ourselves.
Why it is closing. Not enough people used it, and honestly, not enough motivation on our side to keep it running. Rather than let it quietly rot, we are shutting it down properly and leaving this page behind.
Thank you. To everyone who signed up and tried it, everyone who came by our booths, stood in front of a camera for a stranger, and laughed at the avatar that came out: thank you. Those were the fun moments, and they were the whole point.
Every avatar and booth session that was ever completed is below. The 3D models and photos are static files and stay downloadable for as long as this page stays up.
With thanks,
The Anima Machines team
Loading model…
Capabilities
What an Anima could do
From lip-synced speech in 50+ languages to nuanced emotional states, the idea was to make an AI feel alive. Here is what the platform offered while it ran.
Instant Personality
Configure how the avatar sounded, acted, and reacted. Dozens of voice presets, or bring your own.
Multilingual
50+ languages with synced lip movements. The avatar spoke the world's languages, naturally.
Behavior Engine
Triggers for “happy,” “thinking,” or “confused” states. The avatar responded emotionally in real time.
LLM Agnostic
Connected to OpenAI, Claude, Gemini, or a local model. Anima played nicely with any brain you chose.
Real-Time Streaming
Sub-200ms latency from text to animated speech, powered by a custom WebGL engine. The avatar responded faster than a blink.
What people made
4 studio avatars and 5 Augmented Humans 2026 booth sessions. Click any card to spin it around in 3D, or use the icons to download the model and photo.
By the numbers
A frozen snapshot of everything that ever happened on the platform — 11 signups, 4 with something to download. Small numbers, honestly reported.
Character-creation funnel
Started a character
Uploaded a photo
Generated a body render
Fully rigged & textured
Persona mode switches
195 total look switches — visitors kept flipping between all three.
Booth-goer demographics
Gender
Age bracket
5 booth-goers scanned in during the live demo.
Self-reported mood (booth)
Averages across 5 completed booth sessions, 1–10 scale.