HiKid is a free, open-source desktop application designed to help children in non-English speaking countries practice English speaking and listening. It provides an engaging, offline, and privacy-safe environment where kids can converse with an AI, listen to stories, and play word games, all processed locally on their machine.

How It Works

The project implements a full local voice conversation pipeline. User audio is captured and processed by SoX, then fed into a local Automatic Speech Recognition (ASR) server (Qwen3-ASR-0.6B) for transcription. A local Large Language Model (LLM) runtime (Ollama, defaulting to qwen3:0.6b) generates responses, which are synthesized into speech by a local Text-to-Speech (TTS) server (kitten-tts-server) and played back via SoX. This architecture ensures all AI processing and data remain offline, prioritizing user privacy and accessibility.

Quick Start & Requirements

• Installation: Clone the repository, then run npm install.
• Prerequisites: macOS 12.0+ (Apple Silicon/Intel), Node.js >= 20.
• External Dependencies: Requires separate installation of SoX, ASR/TTS servers, and model files; details are in INSTALL.md.
• Development: Use npm run dev for hot-reloading.
• Build: npm run build (all platforms) or npm run build:mac (macOS only).

Highlighted Details

• Offline Operation: All core functionalities—recording, speech recognition, synthesis, and dialogue—operate entirely locally without internet access.
• Privacy-Focused: User conversations, voice data, and AI models are processed exclusively on the user's machine, ensuring no data is uploaded.
• Interactive Learning: Designed for natural conversation; the AI is patient with varying speech speeds and simplicity.
• Kid-Friendly UI: Features a cartoon-like interface intended to appeal to children.

Maintenance & Community

Contributions are welcomed via issues and pull requests. The dev branch is for active development, while main serves as the stable branch. No specific community channels (e.g., Discord, Slack) are listed.

Licensing & Compatibility

• License: MIT.
• Restrictions: Explicitly prohibited for commercial use, resale, or any profit-making activities.

Limitations & Caveats

The application is currently macOS-exclusive; Windows and Linux support are planned but depend on community contributions. Its use is restricted to personal learning, research, and non-commercial demonstration purposes.