Developers · Speech API

The Arabic speech API, built for builders.

One endpoint for the engine under Intella Studio: more than 25 dialects, batch or live, in the cloud or inside your own network.

Transcribe an audio file
# POST an audio file, get dialect-aware text back
curl https://api.intella.me/v1/transcribe \
  -H "Authorization: Bearer $INTELLA_API_KEY" \
  -F "audio=@call.wav" \
  -F "dialect=auto"
Response
{
  "text": "كيف أجدد رخصة القيادة إلكترونياً؟",
  "dialect": "hijazi",
  "confidence": 0.9573,
  "words": [ /* timestamped tokens */ ]
}

Everything you need to ship voice

  • 95.73% accuracy across 25+ Arabic dialects — no MSA translation layer.
  • Real-time streaming over WebSocket for live captions and agents.
  • Clean REST API with predictable JSON and word-level timestamps.
  • Noise-robust models tuned for real call-center and field audio.
  • Cloud, on-premises or hybrid — your audio never leaves your firewalls.
  • SDKs & webhooks to drop into your existing stack in days.

API surface

A small, predictable surface.

POST/v1/transcribeBatch transcription for audio files, dialect auto-detected.
WS/v1/streamLow-latency streaming transcription for live audio.
POST/v1/summarizeSummaries and topics from a transcript or recording.
GET/v1/dialectsList supported dialects and model options.

Illustrative — full reference and keys are shared on access approval. On-premise builds expose the same interface inside your network.

Get your API key and a sandbox.

Tell us your use case and dialects, and we'll set you up with access and a technical sandbox — cloud, on-premises or hybrid.