ASR · Speech to text
Make speech searchable and actionable.
We train and adapt recognition systems for real conversations, local accents, specialized vocabulary, and the languages people actually use.
Duran · Noongil LLC
AI that listens, speaks, and thinks in your language.
Duran works on speech and language AI for languages that mainstream systems often overlook. Our research spans speech recognition, natural voice synthesis, and small language models. We also build products that bring those capabilities to people and organizations.
The research
Three model families, one language stack.
TTS · Text to speech
Give language a natural voice.
We develop speech synthesis for expressive, useful voices in low-resource languages, beginning with Mongolian.
SLM · Small language models
Reason close to the work.
We adapt smaller models to an organization's documents, terminology, and workflows, with deployment choices shaped by its infrastructure and data requirements.
Sovereign AI
Built for your environment.
For organizations that need control over infrastructure and data residency, we work on private and on-premise deployments, fine-tuning and integrating models around their languages and workflows. The exact architecture and data-handling commitments belong in each partnership agreement.
A Duran product
Duran mobile assistant
Our consumer voice assistant is a separate product experience: talk naturally, ask questions, explore content, create images, and manage reminders. Unlike a private on-premise deployment, the mobile app uses cloud services to process requests. Its account, privacy, support, and terms information is grouped in the mobile-app section.