Voices From The Past

Inspiration

History has always been told through books and lectures — static, one-directional, and passive. I wanted to change that. What if you could actually talk to the people who shaped our world? Ask Mahatma Gandhi about nonviolence. Question Einstein about relativity. Hear from Khusrau I Anushirvan about his alliance with the Turks. That spark of curiosity became Voices From The Past.

What it does

Voices From The Past is an immersive time machine web application that connects users with historical figures across all of human history.

• Enter any century (positive for AD, negative for BC)
• Browse AI-generated historical figures from that era, grouped by region
• Start a real conversation — the character only knows what existed during their lifetime
• Available in 4 languages: Uzbek, Russian, English, Chinese
• Prophets are shown with respect — chat is disabled, with the message: "This prophet is currently unavailable. In conversation with the Creator..."

How I built it

Built entirely using MeDo — a zero-code AI application platform. I used natural language prompts to design every screen, animation, and AI behavior. The entire application — from the Matrix-style terminal UI to the typewriter chat effect — was generated through multi-turn conversations with MeDo.

Key MeDo features used:

• Full-stack app generation from natural language descriptions
• Edge Functions for AI character generation
• Multi-turn iterative refinement of UI and logic
• One-click deployment to public URL

External integrations:

• Wikipedia REST API for character portrait images
• Web Audio API for keyboard typing sounds
• Browser SpeechSynthesis API for text-to-speech
• Web Speech API for voice input

Challenges I ran into

• Iframe limitations — some websites block embedding; worked around using embed URLs and Google search
• Edge Function timeouts — character list loading sometimes failed; solved with retry logic
• Typewriter loop bug — text was repeating infinitely; fixed with proper animation state management
• Prophet sensitivity — handling religious figures respectfully required careful prompt engineering
• Time boundary enforcement — ensuring AI never reveals future knowledge to historical characters

Accomplishments I'm proud of

• Created a fully functional time machine in hours, not months
• Immersive hacker terminal UI with Matrix animations, portal effects, and typewriter responses
• Multilingual support across 4 languages
• Culturally sensitive handling of religious figures
• Wikipedia-powered real portrait images for historical characters

What I learned

• MeDo can generate production-quality applications from well-structured prompts
• Prompt engineering is the new programming
• System prompts for AI characters require careful time boundary rules to stay historically accurate
• Zero-code doesn't mean zero thinking — architecture and logic still matter

What's next for Voices From The Past

• Voice output matching character gender and era
• Future century analysis with AI speculation mode
• More languages and regional historical figures
• Collaborative mode — multiple users in same time period
• Educational curriculum integration for schools

Built With

• edge
• medo
• react
• typescript
• web-audio-api
• web-speech-api
• wikipedia-api