Saj Speak
Open Saj Link
Voice capability of Saj Sense

Voice AI that never leaves the device.

On-device voice intelligence — transcription, synthesis, wake-word, and more. From wake-word detection to speech synthesis, every engine runs on the device with zero cloud dependency. Arabic, Hindi and Urdu are built into the foundation alongside English.

How to get Saj Speak Record wake-word samples

Built in Rust by Sajdak Group Holdings W.L.L. Hundreds of tests passing. Patent portfolio filed.

Wake words, one per language
  • Hey SajEnglish
  • يا ساجArabic
  • हे साजHindi
  • ہے ساجUrdu
Wake-word inference
<1 ms
Wake-word model size
<1 MB
Trained models
70+
Languages: English, Arabic, Hindi, Urdu
4

Why Saj Speak

Voice AI that respects your users.

Three principles, no compromise.

  • Private by design

    Your voice never leaves your device. No cloud processing, no data retention, no telemetry. GDPR, HIPAA, and NESA compliant by architecture — not by policy.

  • Arabic-first

    Not a translation. Not an afterthought. Arabic, Hindi, and Urdu are built into the foundation alongside English.

  • Every platform

    iOS, Android, Linux, Windows, macOS, Raspberry Pi, WASM, MCU. A Rust core with native bindings — deploy to phones, speakers, embedded devices, and browsers.

Engines

Every engine runs on the device.

No cloud round-trip. Ship voice features that work offline, everywhere. Each engine is its own Rust crate.

Listen

  • Wake word saj-wake

    Custom keyword detection. Lightweight neural architecture, under 1 MB models. English, Arabic, Hindi, Urdu.

  • Voice activity saj-detect

    Neural voice activity detection with sub-millisecond latency.

  • Streaming speech-to-text saj-listen

    Real-time transcription. English, Arabic, Hindi, Urdu.

  • Batch speech-to-text saj-scribe

    File transcription with a split encoder and decoder.

  • Speech intent saj-intent

    Direct intent extraction. Voice commands to actions, skipping the text. English, Arabic.

Speak

  • Text-to-speech saj-speak-tts

    Natural on-device synthesis. English, Arabic, Hindi.

  • Voice cloning saj-clone

    Clone a voice from samples, on-device. English, Arabic.

Who is speaking

  • Speaker ID saj-voice

    Speaker embeddings for verification and identification.

  • Diarization saj-who

    Real-time speaker segmentation. Who spoke when.

  • Noise suppression saj-clean

    Neural noise removal. Lightweight.

Connect

  • Pipeline saj-pipeline

    Unified voice interaction: wake, voice activity, speech-to-text, intent, text-to-speech.

  • Neural codec saj-codec

    Neural speech codec with quality profiles from 1.2 to 6.0 kbps. Real-time on CPU.

  • Multimodal sensing saj-sense

    Unified token streams with per-modality encryption.

  • Encrypted comms saj-link

    End-to-end encrypted messaging. Sealed sender.

  • Self-hosted server saj-server

    REST and WebSocket, with an OpenAI- and Deepgram-compatible API. Multi-tenant. Docker.

Native Arabic

Arabic in the foundation, dialect by dialect.

420M+ Arabic speakers. 1.09 billion Arabic, Hindi, and Urdu speakers combined.

DialectCoverageStatus
الفصحى
Modern Standard		 Wake word, STT, TTS, intent. The formal register of 420M+ speakers. Supported
الخليجي
Gulf		 Wake word, STT, TTS. Tuned for UAE, Saudi, Bahrain, Kuwait, Qatar, Oman. Supported
المصري
Egyptian		 100M+ speakers. The most widely understood Arabic dialect worldwide. Planned
الشامي
Levantine		 Syria, Lebanon, Jordan, Palestine. 30M+ speakers. Planned

Arabic wake words available

Ya Saj يا ساج Ya Bella يا بيلا Ok Saj اوكي ساج Custom

Enterprise MENA voice AI

Banks, telcos, and government agencies across the Gulf need voice AI that processes Arabic on-device — with dialect awareness, sovereign data processing, and NESA compliance built in.

Contact enterprise sales

Access

How to get Saj Speak.

Saj Link is the application you use every day. Saj Speak, Saj See, and Saj Codec are internal capability modules that surface through Saj Sense Business+, the hosted developer API, and OEM/Embedded engagements — never sold as separate brands.

  • Saj Sense Business+

    Saj Speak surfaces in the Saj Sense Business+ tier.

  • Hosted developer API

    Saj Speak is also available through the hosted developer API.

  • OEM / Embedded

    Saj Speak inside your own device or product, through an OEM/Embedded engagement.

See Saj Sense pricing Contact sales

On-device

Why on-device?

Cloud voice APIs charge per minute, leak data, and add latency. On-device fixes all three.

  • Total privacy

    Zero audio data leaves the device. No cloud processing. No data retention.

  • Sub-millisecond wake word

    Sub-millisecond wake-word inference against a 200–500 ms cloud round-trip. Real-time voice interaction without waiting for the network.

  • No per-minute charges

    No audio-hour billing. Your models run on your hardware.