Inspiration

LegacyTree was inspired by the realization that when our loved ones pass away, we often lose more than just their presence we lose their stories, their voice, and their lived wisdom.

My grandmother once told me a story about how she escaped a riot as a child. It was powerful, heartbreaking, and full of resilience. She only told it once and we never recorded it. That memory, like so many others, is now gone forever.

What it does

LegacyTree is a conversational, AI-powered storytelling platform that helps families preserve life stories in meaningful and interactive ways. It allows elders and loved ones to share their memories, which are then transformed into a rich digital memoir.

✨ Key Features:

Conversational Story Recording: AI interviews elders using thoughtful, guided prompts.

AI Summarization & Theming: Stories are automatically titled, tagged with themes, and structured into chapters.

Interactive Memory Map: Pins stories to real-world locations based on where events took place.

Photo & Voice Artifact Upload: Users can add family photos, audio clips, and personal documents.

How I built it

Frontend: Built using Streamlit for a fast, accessible user interface.

Backend: Developed with FastAPI, including REST endpoints for:

/conversation: AI-guided interviews using GPT-4

/transcribe: Speech-to-text via OpenAI Whisper

/summarize: Story summaries and themes

/illustrate: AI image generation (DALL·E)

/map: Geolocation of memories using Google Maps API

Database: SQLite (legacytree.db) to store stories and metadata.

AI Tools:

OpenAI GPT-4 for storytelling and conversation

OpenAI Whisper for transcription

DALL·E for optional illustrations

Prompt engineering for personal tone and question sequencing

Challenges I ran into

Designing AI interactions that feel natural and emotionally sensitive

Managing asynchronous, stateful conversations across multiple sessions

Handling multimedia uploads and mapping with limited resources

Scoping a deeply emotional and ambitious idea into a hackathon-sized MVP

Accomplishments that I'm proud of

Building a complete, working MVP as a solo participant

Successfully integrating OpenAI's GPT and Whisper APIs into a storytelling pipeline

Creating a seamless user experience where voice stories are transformed into AI-written chapters and mapped to real places

Designing a tool that blends technology with empathy, something people can emotionally connect with

What I learned

How to combine technical tools with human emotion

Designing AI to listen, not just respond

Building full-stack applications solo under time pressure

That meaningful tech doesn’t have to be flashy, it just needs to connect with people

What's next for LegacyTree

Voice cloning (opt-in) to preserve how loved ones actually sounded

Emotion-based story timeline showing highs, lows, and major life chapters

Community Memory Circles to crowdsource stories by culture or history

Time capsule releases: Unlock memories on future birthdays or anniversaries

Mobile version for easier story capture from anywhere

Built With

Share this project:

Updates