Lumina Twin

Lumina Twin is a lifelong AI companion that learns from your experiences, conversations, and decisions to provide personalized guidance while preserving your knowledge and legacy.

Comment

Inspiration

Lumina Twin was inspired by a very personal question: what if I could still talk to my grandfather?

Not only to ask big life questions, but simple human questions too: what was his favorite color, what did he care about, what would he think about something happening today, or what advice would he give me as his grandchild?

Most AI chatbots answer from general internet knowledge. Lumina Twin is built from the opposite idea. A Twin should grow from the user’s own approved memories, habits, voice, photos, writing, reflections, browsing signals, daily answers, and personal patterns.

The goal is not to create a fake person. The goal is to preserve a consent-based digital version of how someone thinks, speaks, remembers, and responds, so their family can stay connected to their personality, values, and life experience.

What It Does

Lumina Twin lets a user build a private digital Twin over time.

With the user’s permission, Lumina Twin collects approved personal data from different sources, including the web dashboard, Android Collector app, Chrome extension, Telegram connector, photos, voice notes, written memories, daily questions, and other consent-based inputs.

The system does not simply store this information like a database. It analyzes the data to understand the user’s personality, preferences, values, habits, interests, emotional tone, memories, relationships, and decision patterns.

For example, if a user often buys black clothes, saves black shirts, or repeatedly shows preference for black items, Lumina Twin can infer that black may be one of their preferred colors. If the user often reads about technology, family, markets, travel, or politics, Lumina Twin uses those repeated signals to understand what the user cares about.

Trusted people can later chat with the Twin through the web or Android Chat app using access keys or QR codes controlled by the owner. The owner can revoke access, pause collection, delete records, export data, restrict sensitive topics, and decide what the Twin is allowed to answer.

How We Use Data

Lumina Twin is not designed to be a simple storage app or a search engine. The purpose of the data is to understand the person behind the records.

Every account begins like a blank mind. Over time, as the user approves more data, the Twin becomes more complete. It learns from repeated patterns, not only from one isolated memory.

The data helps Lumina Twin build a personality model, including:

  • What the user likes and dislikes
  • What topics they care about
  • Their values and priorities
  • Their communication style
  • Their emotional tone
  • Their habits and routines
  • Their memories and relationships
  • Their decision-making patterns
  • Their likely preferences when direct evidence is not available

The long-term goal is for the Twin to think and respond as close as possible to the user, based on approved evidence. If the user passes away, their Twin can continue to answer trusted family members in a way that feels close to how the user may have thought, spoken, and responded.

This is why Lumina Twin focuses on personality analysis, not just saved information.

How We Built It

We built Lumina Twin as a full product foundation, not only a demo.

The platform includes:

  • A web dashboard for onboarding, billing, memory upload, data review, privacy controls, guest access keys, QR access, and Twin chat.
  • An Android Collector app for memories, voice, photos, location notes, Telegram connection, permissions, billing, and voice profile consent.
  • An Android Chat app for chatting with saved Twins, QR/key access, voice replies, room-style conversations, reporting, and local chat controls.
  • A Chrome extension that automatically collects approved browser context after the user connects a revocable token.
  • A Telegram connector for forwarding approved messages, notes, photos, and voice content into the Twin memory system.
  • A backend with account management, authentication, email verification, subscription billing, secure secrets, data export, deletion controls, and privacy safeguards.

The main design principle is:

$$ Better\ Twin = Approved\ Data + Personality\ Analysis + User\ Control $$

We designed Lumina Twin around consent. The user should always know what is being collected, why it is being collected, and how to stop it. They can pause collection, revoke permissions, delete data, export records, and control who can speak with their Twin.

Language and Conversation

Lumina Twin is designed to respect the language of the conversation.

If a family member asks a question in Spanish, the Twin should answer naturally in Spanish. If they ask in English, it should answer in English. The experience should feel like speaking with the person, not using a translation tool.

The Twin should also answer naturally, not like a search engine. If the user asks, “What is your favorite color?” and the data strongly suggests the person prefers black, the Twin should answer directly, such as: “I think my favorite color is black. I seem to choose black clothes often, especially shirts.”

The goal is for the Twin to feel human, direct, and emotionally believable while still being based on approved data.

What We Learned

We learned that building a digital Twin is not only an AI problem. It is also a trust, privacy, consent, UX, billing, and emotional design problem.

Some of the biggest lessons were:

  • Users must always know what data is collected.
  • Users must be able to pause, revoke, delete, and export their data.
  • A Twin must answer like a person, not like a search tool.
  • The system should analyze patterns over time, not just retrieve old records.
  • Language consistency is important for emotional trust.
  • Raw sources should not leak into chat replies.
  • Voice cloning needs explicit consent and clear control.
  • Older or non-technical users need a simple experience.
  • Access keys and QR codes make family sharing easier.
  • The product must protect sensitive topics like health, legal issues, finances, relationships, religion, politics, trauma, and private family conflict.

Challenges

The hardest challenge was making the Twin feel human while staying honest.

If the system has strong evidence, it should answer directly. If it does not have enough evidence, it should still respond naturally and carefully, instead of sounding robotic or pretending to know something it does not know.

Another major challenge was privacy. Lumina Twin collects deeply personal context, so we had to design controls for consent, data deletion, guest access, topic restrictions, subscription status, inactivity, export, and account removal.

We also faced technical challenges across multiple platforms:

  • Keeping the web app, Android Collector app, Android Chat app, Chrome extension, and Telegram connector connected to the same backend.
  • Making subscription and free trial rules work across apps.
  • Handling voice samples and synthetic voice consent.
  • Preventing raw browser text or irrelevant data from appearing in chat answers.
  • Supporting multilingual conversations.
  • Making the apps simple enough for real users, not only for technical users.
  • Building a product that feels emotional and useful, while still respecting safety and privacy.

What’s Next

Next, we plan to improve the Brain layer so each Twin grows more deeply over time.

The goal is to move from storing memories to building a stronger personality model: preferences, beliefs, emotional tone, communication style, life priorities, relationships, habits, values, and confidence levels.

Planned improvements include:

  • Smarter personality analysis over time.
  • More natural family-style conversations.
  • Better multilingual replies.
  • More automatic consent-based connectors.
  • Stronger voice profile and voice reply controls.
  • iOS Collector and iOS Chat apps.
  • Trusted person continuation flow for legacy access.
  • More advanced privacy and topic restriction settings.
  • Better onboarding for families, immigrants, aging parents, and people who want to preserve personal memory.
  • A stronger system for asking thoughtful daily questions based on what the Twin does not yet know.

Lumina Twin is built around one belief: people should be able to preserve their digital mind with consent, dignity, and control.

Built With

Share this project:

Updates