Quickstart
First synthesised audio in five minutes.
API reference
Every endpoint, parameter, and error.
Two ways in
Speech API
Turn text into speech over HTTP or WebSocket. 14 voices across Najdi, Gulf, MSA, Egyptian, and English, with telephony-ready output.
Voice agents
Complete conversational agents that answer calls, take action in your systems, and hand your team a bilingual record.
Start here
Voices
All 14 voices by dialect, style, and age.
Models
Sada for live conversation, Nabra for volume.
Streaming
First audio in ~200 ms instead of ~1.7 s.
Writing instructions
The highest-leverage thing you control on an agent.
Testing
Find the failures before your callers do.
Troubleshooting
Common problems and the fix for each.
What an agent does on a call
Answers
Picks up, greets the caller in your configured dialect, starts listening.
Understands
Transcribes and interprets in real time, including Arabic–English code-switching.
Acts
Classifies the request and calls your systems — logging a ticket, checking an order.
Reports
Leaves a transcript and a bilingual summary, plus any records it created.
Where teams deploy them
Service desks
Take issue reports, classify them, and log tickets without a human triaging the queue.
Customer lines
Answer product questions, quote prices, and capture orders on the first call.
Facility and device control
Let callers operate equipment by voice, with the command executed live.
Custom workflows
Describe the job in plain language and run an agent shaped around it.
New to voice agents? Read Core concepts first — it is ten minutes and every other page assumes it.

