Koveh Translate
I built a translation service that works with multiple AI models. You choose what fits your task.
Local model Helsinki NLP translates between Russian, English, and German. Fast and private — everything stays on my servers.
External models (Google, OpenAI, DeepL) handle 25 languages with higher precision. Good for complex texts and rare language pairs.
Voice generation turns any translation into speech. I made this for podcast creation, but it works for any audio content. You control:
- Voice model and speaker
- Speech style and tone
- Custom prompts for context
The interface adapts to phones and desktops. No separate mobile app needed.
API for Business
Behind the scenes runs api.koveh.com — a unified API hub for translations, AI, and automation tools. Instead of managing dozens of different APIs, companies connect to one endpoint and access everything.
Register at koveh.com/login to explore the full service catalog. New tools added regularly.
Technical Architecture
The system runs on microservices architecture:
- Docker containers for each service
- Kubernetes orchestration for scaling
- RabbitMQ for task queuing and load balancing
This setup handles high loads without service dependencies. When one component updates or fails, others keep running. Each translation request gets processed by the most suitable available instance.
The modular design lets me add new models and features without touching existing code. That's how I keep the service stable while constantly improving it.
