*This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend
What I Built
I built Speech Translator, a privacy-first, offline-capable web application designed to bridge communication gaps for a loved one. Specifically, I built this for an elderly family member who struggles with language barriers during healthcare visits and daily conversations.
Speech Translator bridges communication gaps for loved ones. Built with HTML and JavaScript, it combines browser speech APIs with local open-weight AI for 100% private, offline-ready, and cost-free translation with zero cloud tracking, ensuring your conversations remain secure and entirely on your device. It eliminates the friction of traditional translation apps by offering an intuitive, high-contrast interface with real-time speech-to-text and text-to-speech output.
Demo
- Live Website: https://amritakaur1717-collab.github.io/Speech-Translator/
- GitHub Repository: https://github.com/amritakaur1717-collab/Speech-Translator.git
Code
You can explore the clean, lightweight codebase directly through the GitHub repository link above. The application is built using a minimal structure containing HTML and JavaScript files.
How I Built It
The application is structured into a lightweight, modular setup using HTML and JavaScript (with no external CSS dependencies).
- Speech Processing: It harnesses native browser Speech Recognition and Speech Synthesis APIs to handle voice input transcription and clear audio playback for translated text.
- Local AI Inference: For intelligent, context-aware translations, the app connects seamlessly to local open-weight AI models (such as Llama or Mistral running via Ollama).
- Architecture: The frontend handles real-time audio event listeners and updates the DOM dynamically while communicating with the local API backend for high-speed linguistic conversion.
Why Does Open Innovation Matter?
Open innovation is the absolute backbone of this project because of two non-negotiable requirements: data privacy and accessibility.
- Absolute Privacy: Personal health discussions and everyday family chats are deeply sensitive. A closed, proprietary cloud API would send voice data and translations to third-party servers where they could be logged or tracked. Open-source local inference ensures that data never leaves the local device.
- Cost-Free & Offline Resilience: Relying on commercial cloud translation or LLM APIs introduces recurring subscription fees or rate limits. Open-weight models running locally allow unlimited, free translations even when an internet connection is unstable or entirely unavailable.
My Agent Session
- Agent Development Session: [Link to your DevRelay session or session logs, if applicable]
Prize Categories
- Open Source AI Integration
- Privacy & Security Innovation
- Best Accessibility Tool
Top comments (0)