The lineup

Two production-ready Uzbek speech AI models.

Rumi synthesizes the most natural Uzbek speech. Nava transcribes the most accurate conversational Uzbek. Both built on the same verified Biruniy Gold dataset.

Available models

Speech synthesis and transcription for production.

Choose Rumi for natural Uzbek voice synthesis or Nava for accurate conversational transcription. Open each model to compare its capabilities, deployment profile, and best-fit use cases.

Rumi

Text to speech

Production

The most natural Uzbek voice synthesis. Streaming, low-latency, 10 dialect voices.

Rumi generates the most natural-sounding Uzbek speech of any TTS system. Trained on 200+ hours of human-verified conversational Uzbek audio from the Biruniy Gold dataset, every training segment carries emotion labels, sentence-complete markers, and prosody tags — so the model learns how real Uzbek speakers breathe, pause, and emphasize, not just how they pronounce individual words.

Trained on 200+ hrs of human-verified Biruniy Gold audio
Emotion labels + prosody tags for natural intonation
Streaming TTS for conversational voice agents

View details

Try Rumi License

Nava

Speech to text

Production

8–9.1% WER on conversational Uzbek. Streaming, 10 dialects.

Nava transcribes conversational Uzbek at 8–9.1% WER — the lowest of any publicly available Uzbek ASR. It supports streaming partial results, semantic endpointing, 10 dialect regions, and native code-switching between Uzbek and Russian.

Ranked #1 on real-world conversational Uzbek accuracy
Built-in turn detection and semantic endpointing
Streaming partial results for real-time voice agents

View details

Try Nava License

License any model
for your use case.

Buy once, run on your own infrastructure. No per-call fees, no API keys, no usage caps.

Request a license Talk to sales

Two production-ready Uzbek speech AI models.

Speech synthesis and transcription for production.

Rumi

Nava

License any modelfor your use case.

License any model
for your use case.