clone.arnao.ai · concept studio

A recorder that learns you and becomes you — done responsibly.

Feed it camera, mic, speech, video, pictures, files, URLs and agents; it builds a persona model that looks, sounds, and acts like the source — in a web interface and as an agent. Quality vs. latency is a slider, but realism is king. Worked example: a dignified legacy clone of a grandfather.

Responsibility is the product, not the fine print. Every concept foregrounds consent capture, disclosure-by-default ("AI clone"), provenance on every claim, scoped agent permissions, and a kill switch. A clone from the RAI person that can't show its receipts is a defect.
Multimodal ingestion Persona model Web + agent surfaces Quality ⇄ latency PRD → Fable PRD v2 · full product →
Likeness Studio · the Builder
Google-auth in, record yourself (camera/mic), drop speech/video/pics/files/URLs/agents — six fidelity meters (face · voice · cadence · knowledge · mannerisms · values) rise as it learns you. Consent + memorial-provenance gate; quality/latency slider. "Teach it to be you."
ingestionfidelity metersconsent gate
Open concept →
Presence · Face to Face
Talking with a realized clone — the dignified memorial example. A breathing avatar, an always-on "AI clone · family-authorized · consent on file" ribbon, and hover-provenance tracing each line to its source (including honest "inference, not his words").
avatar + voicedisclosure ribbonprovenance
Open concept →
The Understudy · Clone-as-Agent
The clone acting autonomously in your voice/values — wrapped in governance: scoped permissions (draft ✓ / may NOT transact / must disclose), a live audit log stamped "disclosed as clone," and an instant kill-switch. Governance lives in the harness.
scoped permissionsaudit logkill-switch
Open concept →
The Persona Model
The interpretability view: the seven facets it has learned (face, voice, cadence, knowledge, mannerisms, values, memories), each with a confidence level and provenance — and a "correct this trait" flow ("I'd never say that" → the model updates). See and edit your clone's model of you.
interpretabilityeditableprovenance
Open concept →
Digital Twin · Simulator 3.0
Not a persona you talk to — a twin you SIMULATE with. Enter a real decision; your twin forks into 5 parallel futures played out in your first-person voice, then reports which way "you" lean and why (traced to your persona traits). Twin-vs-twin stress tests current-you against past-you. A labeled simulation, never a prediction.
digital twinforked futurestwin vs twinsimulator 3.0
Open concept →
The Living Chat-Head
An ambient, breathing avatar of the person that lives at your screen's edge — glancing, blinking, present — and grows to full presence when you engage it, then recedes when idle. A companion, not a widget. Always-on "AI clone" badge, hover-provenance, ambient⇄engaged render.
ambient presencedock ⇄ fullcompanion
Open concept →
The Memorial Room
A witnessed, family-authorized space to sit with a departed parent. Honesty is built into the grammar — his real words appear in quotes, teal, linked to a specific letter/voicemail/video; inference is rendered plainly as reconstruction, never faked as a quote. A shared guestbook, candles, "who is present." Healing, not deception.
memorialestate consentwords vs inferencewitnessed
Open concept →
Mirror Mode
You look in; your clone looks back and reflects you live. Slide from exact mirror toward fully-autonomous-you and watch it take over — "mirror me until it's me" (divergence = autonomy × (1 − fidelity)). Face-unlock entry with passkey step-up, an on-screen gesture map, and a keyboard-shortcut cheat sheet.
live mirrortrain ⇄ autonomousface-unlockgestures + keys
Open concept →
Concept mockups — real interactive UI, no live cloning behind them yet. The through-line: responsibility as the feature. Tell me the direction (the PRD argues for the Builder first) and I build the engine — reusing your local Chatterbox voice-clone + the Gia clone skill + your agent fleet. · Built by a PM agent + design agents.