Arabic Speech to Text · STT

Transcription built for real Arabic calls and noise.

Turns spoken Jordanian, Saudi, and Lebanese Arabic into clean text—tuned for phone-line noise, numerals, and mixed English terminology.

3 dialectsTuned for Jordanian, Saudi & Lebanese
AR ⇄ ENHandles mid-sentence English words
8 / 16 kHzTelephony & browser audio preprocessing
REST /v1Standard /v1/audio/transcriptions route

Engineered for noisy real-world audio

01 · Preprocess

Noise cleanup & silence trimming

Normalizes WebM/M4A/WAV inputs and strips dead air so background hum never turns into phantom words.

02 · Recognize

Dialect phrasing & spoken digits

Recognizes everyday Jordanian, Saudi, and Lebanese expressions alongside English product terms.

03 · Route

Feeds agent tools & post-call QA

Passes clean transcripts straight to your workflow tools and post-call analytics.

Speech-to-Text highlights

Order IDs & phone numbers

Reliable digit capture and normalization for order lookups and booking references.

No phantom words on silence

Clean VAD and audio preprocessing prevent background hum from triggering false turns.

Browser & telephony formats

Accepts WAV, MP3, M4A, WebM, and OGG through a single multipart endpoint.

Full audio sovereignty

Run entirely inside your VPC or self-hosted GPU so recordings never leave your perimeter.

REST /v1 transcription example

curl -X POST https://api.muwakilai.com/v1/audio/transcriptions \
  -H "Authorization: Bearer $MUWAKIL_API_KEY" \
  -F "file=@call.wav" \
  -F "language=ar"

Common questions

Which audio file formats are supported?

WAV, WebM, M4A, MP3, and OGG are decoded and normalized automatically.

Can we test it right now without an account?

Yes, switch to the Speech to Text tab on the homepage sandbox and record or test a dialect sample.