# blackrossay

- **Event:** [Meta Global AI Developer Hackathon](https://cerebralvalley.ai/e/meta-ai-developer-hackathon)
- **When:** Oct 26, 12:00 PM – Nov 4, 9:00 PM (EST)
- **Where:** Online
- **Team:** [Ashley Chamboko](https://cerebralvalley.ai/u/blackrossay)
- **GitHub:** https://github.com/blackrossay/twinverse-ai
- **Demo video:** https://youtu.be/1FM_1eKt6Y4
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3886

Ask once. Hear every side of yourself. Clone your voice & face once — your twin debates decisions from 3 lenses, talks with memory in your cloned voice, and goes Live as you on YouTube. Built with muse-spark-1.3.

Your voice. Your face. Live.
TWINVERSE AI is a persona platform for streaming creators and busy professionals who can't be present everywhere. Generic AI doesn't know you. We fix that.

You build your Persona once — verified facts are kept strictly separate from confidence-scored hypotheses, with cloned voice + likeness. That one identity powers everything:

1. Twin Council — The core demo in your video
One question, three minds of you in parallel. The Strategist argues go, the Skeptic argues stop, the Counselor weighs the human cost. Every verdict is evidence-scored with deterministic code checks. Synthesis is traceable, reruns are byte-identical with a diff view. You see it in the video: seed personas -> Ask the council -> 3 cards running in parallel -> comparison screen.

2. Talk — Your conversational clone
A multi-turn voice agent that remembers the thread and answers in character. Tap the Voice Orb to speak, it transcribes live, thinks with a shimmer, and answers out loud in your cloned voice and face. All captions included, barge-in supported, typing as fallback. This is voice-first, not chat-first.

3. Live Twin Studio — For creators
Your persona goes live on stage. Heads-up discussion points, real-time chat answers from simulated audience across platforms, Business / Life / Socials modes, and one-click publishing to YouTube via RTMP (stream key masked server-side). What you saw in the clicks: Talk -> Live -> clone on stage deliberating.

Foundation you can trust:
- Facts vs Inferences never mixed (dashed "inferred, not verified" UI)
- Every score traces to evidence
- All model calls server-side, MODEL_API_KEY never hits the browser
- Offline fixtures are clearly labeled [OFFLINE FIXTURE] — never faked as live
- Fully local open-source voice/avatar pipeline (MeloTTS + OpenVoice + MuseTalk) or your own cloud inference service

How we built it:
Frontend: Next.js 15 PWA (installable, mobile bottom-nav, dark-only stage product with glowing glass system), Tailwind, Instrument Serif + Space Grotesk, Apple-mapped springs for motion.
Backend: FastAPI + SQLite, versioned prompt templates shipped in wheel (twinverse/prompts/v1/).
AI: muse-spark-1.3 for extraction, reasoning & evaluation + muse-image-1.0 for identity cards via Meta Model API.
Infra: docker compose up --build runs everything — backend :8000, frontend :3000.

Why this is different:
Not a generic avatar with a script. It's provably you — grounded in your identity, speaking in your voice, live in your face, under your rules. One identity — deliberates, talks, and streams everywhere at once.

Run in 2 mins: docker compose up --build 

Stack: Llama 4 Scout, Llama Image, FastAPI, Next.js PWA, local voice cloning
Repo: https://github.com/blackrossay/twinverse-ai

## More from Meta Global AI Developer Hackathon

- [AuraHealth](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3883)
- [OVN Nexus](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3884)
- [Brisky](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3885)
- [Byte me](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3887)
- [Causal Delta](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3888)
- [AMONS](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3889)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3886. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt