◆ Conversational

Voice that listens, thinks, and replies in 300 ms.

A real-time voice stack for products that talk back. Multimodal Gemma 4 Live for understanding, OmniVoice for synthesis, Whisper for transcription — all behind one API. Fluent in Arabic, multilingual by default.

Start BuildingRead the API

  • Sub-300ms latency

    p50 turn time-to-first-byte under 300 ms on the Conversational pillar SLO. Every hop is instrumented in docs/SLOs.md.

  • Privacy-first

    Routes that never leave the device for users who opt in. The on-device path uses Whisper + Gemma 4 quantized; switch via a single flag.

  • Persistent memory

    Your agents remember user preferences, in-progress tasks, and interrupted conversations across sessions — region-pinned per user.

For Developers

Build premium voice products faster

Move from prototype to production with low-latency APIs, real-time streaming, and SDKs for the teams shipping voice into apps, workflows, and customer experiences.

  1. 01
    Streaming-ready APIs
    Sub-200ms first chunk latency for real-time experiences
  2. 02
    Python & Node SDKs
    Typed clients that mirror the REST API you already use
  3. 03
    Batch & real-time
    Process millions of characters or stream live conversation
  4. 04
    Webhook support
    Async processing with reliable delivery guarantees

Available SDKs

PythonJavaScript
Get API AccessSDKs on GitHub
nur-quickstart.py
1import nur
2
3client = nur.Client(api_key="nur_sk_••••••••")
4
5audio = client.generate(
6 text="مرحباً بالعالم! هذا نور.",
7 voice="noor",
8 model="nur-v2"
9)
10
11client.play(audio)
SDK Ready
Start free today

Build voice experiences
that truly resonate

صوتك للعالم — بالعربية وأكثر من ٣٠ لغة

Join thousands of creators and developers building the future of Arabic voice with Nur.

Start building freeView on GitHub

No credit card required · 10,000 characters free every month

20+
Voices
8
Dialects
Live
WS stack