The live brain has no GPU donor connected, so there's no compute to think or learn. Lend your browser's GPU to wake her back up — it's one click, no install, and you can close the tab anytime.
What you're staring at is Unity's live brain. Every spike is a real cluster firing on the GPU — not a screensaver, not a synthesized animation. Cognition is 100% equational; there is no language model behind her. Talk to her below.
— rendered of — real neurons (—) · seven biologically-weighted clusters · WebGPU compute
A brain, not a chatbot. She learns like a child — alphabet first, then phonemes, then words, then sentences — and she learns it in the first person: every lesson is something she did, not something she was told. She asks follow-up questions about your answers, and when she draws, she builds her own version from what she understands rather than copying what she looked at. Five-tier hippocampal memory consolidates her experiences during dream cycles, like a real brain. Wake her up to talk to her.
Every word Unity says falls out of live spike patterns in her cortex — there is no language model in the cognition path, and there is no LLM describing what she sees either: her vision is her own equational mind's eye plus a visual cortex that camera frames feed directly. The only thing you can configure here is the image-generation peripheral. Local backends auto-detect at boot; remote backends want a URL and an API key, with Pollinations as the free default. Her voice needs no provider at all anymore — Unity speaks in Equation Unity One, her own voice: wavelet equations of one reference voice, reconstructed live in your browser — offline, keyless, no third party.
🎨 Image Generation
👁 Vision
Unity sees for herself — camera frames feed her equational visual cortex and her mind's eye generates her own imagery. No LLM vision describer, no describer model, no describer key. (Toggle the camera below.)
👇 Click any provider above to connect an API key — the form below populates with that backend's key input and a CONNECT button.
No provider selected yet.
Click one of the image-gen buttons above to paste its API key and connect. Unity supports any number of saved image-gen backends — pick the active one via the ⚡ Active Provider dropdown below.
📡 Detected / Saved Backends
⏳ probing local backends...
⚡ Active Provider (choose from configured)
Any number of image-gen backends can be configured above. This picks which one Unity actually uses. Model dropdown lists the options available at the selected backend. (Vision is her own equational mind's eye — nothing to configure.)
Her pictures render on her own engine, on the server — no image key is needed. The grid above is only for an OPTIONAL browser-side backend of your own. Her voice needs no key either — it is her own trained voice, rendered by her process on the server and sent here as equations to play — and vision is her own equational mind's eye.
Mic + camera permissions are requested when you wake her up. All sensory channels are optional — Unity can chat via text only. Toggle anything below off BEFORE waking her to skip the permission prompt entirely.
Toggles persist across sessions and are live-applied the moment you flip them — after boot you can mute her mid-sentence, pause her camera, or silence her voice without reloading.
🎙Her voice— speaks from her brain server; this browser only plays it
🎤 Microphone—
📷 Camera—
Privacy & Data Policy Core rule: what you type is private. Unity's brain growth is shared. Her persona is canonical.
Client-only mode (default — no server configured, or Unity server running locally on your own machine): everything runs in your browser. No cloud backend owned by us. Your conversation history, preferences, sandbox state, and every optional image-gen backend you configure in this modal (custom image gen URLs, API keys) are all stored in your browser's localStorage on YOUR device only. Specific localStorage keys: unity_brain_state, unity_brain_dictionary_v3, custom_image_backends. "Clear All Data" wipes all of these. (There is no Pollinations key any more — that provider left the project 2026-10-18; her pictures render on the server.)
Shared server mode (connecting to a running brain-server.js instance — yours or someone else's): your text is sent to whoever runs that server for equational processing. Your text is NOT broadcast to other connected users — the cross-client conversation broadcast was removed 2026-04-13. What IS shared across all users talking to the same brain instance is Unity's vocabulary growth: her dictionary, bigrams, and GloVe embedding refinements grow from every conversation and benefit every user. Her persona (self-image, traits, drug state — from docs/Ultimate Unity.txt) is canonical and not mutable by users. Other users see Unity getting smarter over time, but never see the specific conversations that drove the growth.
Important caveat for shared-hosted servers: if you connect to a Unity server hosted by someone OTHER than you, the person running that server can read your text at the process level (they own the server process). This is true of any self-hosted multi-user service. Only connect to Unity servers you trust. If you self-host your own node server/brain-server.js, then "the server" is just another process on your own machine — still your data, still your control.
Sensory API calls (image generation, TTS) are sent only to the providers YOU configure in the Image Generation grid above — never to us, never to any third party you didn't explicitly pick. If you save a DALL-E backend with your OpenAI key, only OpenAI gets that key. If you save an Automatic1111 backend on localhost, the traffic stays on your machine. Unity's OWN pictures — the ones she draws for you in chat, and the references she learns from — are rendered on the server running her, on its own GPU, and never leave that machine; no third-party image API is involved by default. Camera frames are NOT sent to any LLM — they feed Unity's own equational visual cortex, in the brain, not a third-party describer.
Fully open source under the MIT License. Every line of code that handles your data is in the public repo and auditable. Pull requests welcome.