LegacyVoice

Inspiration

We say the most important things to the people we love way too rarely. And when we finally do, it's usually just text — a message, a card, something that gets read once and forgotten. We wanted to build something that let people say the things they mean, and have it actually sound like they meant it.

What it does

LegacyVoice lets you write a memory or message to someone you love — who it's for, who it's from, and the memory itself, in your own words. Once you create your letter, it's read back to you in a warm, natural voice using ElevenLabs text-to-speech. Every letter you keep is saved to your own keepsake shelf, ready to be heard again anytime.

How we built it

The frontend and app logic are built on Base44. For narration, we integrated ElevenLabs' Text-to-Speech API through a custom backend function — the letter text is sent securely, converted to speech server-side, and streamed back as audio. We designed a custom voice through ElevenLabs' Voice Design tool to get a warm, natural tone that fit the emotional purpose of the app, rather than using a generic default voice.

Challenges we ran into

Getting real-time text-to-speech working smoothly took real trial and error — we hit payment-tier restrictions on ElevenLabs' shared voice library, had to design and train a custom voice from scratch to get around free-tier limitations, and debugged a race condition where the first playback click would get interrupted before audio finished loading. We also had to be deliberate about credit usage once we realized how limited free-tier ElevenLabs credits are for a project this audio-heavy.

What we learned

We learned a lot about working with third-party AI APIs in a real product context — handling API errors gracefully, managing secrets securely, and designing around real constraints like tiered access and usage limits, not just the happy path. We also learned how much a single well-chosen detail (like a custom-designed voice, or a small animation like the wax seal) can carry the emotional weight of a product far more than a feature list ever could.

What's next for LegacyVoice

We'd like to add word-synced highlighting that works with ElevenLabs' narration (not just the browser voice fallback), support for multiple saved voices per user, and a way to schedule a letter to be delivered or unlocked on a specific future date — turning it into a true time capsule.

Built With

• ai
• and-hear-it-read-back-to-you-?-a-heartfelt-letter
• base44
• elevenlabs
• javascript
• preserved-forever
• react
• text-to-speech
• voice-synthesis
• web-app