Max Translator: Intelligent AI Voice Translator

Objective

Create a mobile web application for natural communication between people in different languages, solving the problem of "machine" literalism and loss of meaning in complex dialogues.

Solution

A hybrid system was developed, integrating advanced cloud technologies. Key modules include:

• Multi-LLM Engine: Connection to Gemini, OpenAI, Groq, and Mistral through a Cloudflare Workers layer for maximum speed and translation quality.
• Contextual Translation (The Scene): Ability to set a situation (e.g., "at the hospital" or "rental negotiations"), allowing the AI to choose specific terms.
• Double Control System: Automatic "Back Translation" with semantic analysis of meaning matching (in percentages).
• Vision AI: Module for recognizing and discussing images (menus, documents, road signs).

Technical Stack

The application is built on a monorepo architecture: React 19, Tailwind CSS, Firebase (Auth/Firestore). For sound processing, the Web Audio API with a custom silence detector is used. Billing is implemented via Firestore atomic transactions with per-second resource charging.

Check it out

You can see the service in action and use its services on the website: tr.ai-robot.us.