preludeofme/heard-again

Heard Again is a platform for digitizing and preserving family stories. It combines a web app for organizing family history with an AI voice service that can turn text into speech using a cloned version of someone's voice, so a written memory can be heard back in a familiar voice. The main web application, built with Next.js and React, is where most of the work happens. It includes a visual family tree with relationship mapping, tools to upload, edit, and organize stories, a timeline view of family events, and the ability to group stories into themed collections. There is also support for importing and exporting genealogy data in the GEDCOM format that many family history tools use, and login with multi factor authentication. The voice part of the project is a separate Python service that clones a voice from a short reference audio sample, or designs a new voice from a plain language description like warm, gentle, or nostalgic. It can then synthesize speech in that voice from any text, and it can run faster with an NVIDIA graphics card, though a card is not required. Because the project has two very different pieces, a Node.js web app and a Python voice service, each has its own separate setup and its own Docker container, and both are tied together by one shared PostgreSQL database and Redis cache. An install script handles checking dependencies, creating configuration files, starting the database and cache in Docker, running database migrations, and building the web app in one go, or you can follow the manual steps in the README to do each part yourself. To run everything you need Node.js, Python, PostgreSQL, Redis, and Docker, plus an NVIDIA GPU if you want faster voice generation. The project is released under the MIT License, which allows free use including commercial use, and it accepts outside contributions through a documented process.