Arabic Dialects · Speech to Text
Lebanese · لبنانيLebanese Arabic
Speech to Text
An STT model that understands Lebanese Arabic with its natural Arabic-French-English code-mixing, serving Beirut, diaspora communities in Latin America, Africa, Europe, and Australia.
dialect: "lebanese"Lebanese voice. Real accuracy.5M+
Native Lebanese Arabic speakers
<16%
WER on Lebanese colloquial speech
<300ms
First-word streaming latency
Trilingual
Arabic-French-English code-mixing
Try it
Speak Lebanese Arabic.
Watch it transcribe.
Record yourself speaking Lebanese Arabic and watch the words appear as you talk. The demo runs Hamsa's full production STT, the same model your API calls will use.
Speech to Text · speech in → text out
No sign-upYour words will appear here.
Press record and speak Arabic or English. The words appear as you talk.
Lebanese Arabic Overview
What is Lebanese Arabic?
Lebanese Arabic is a Levantine dialect with a distinctive Francophone character, spoken by Lebanese communities in Lebanon and across a vast global diaspora. Its natural Arabic-French-English code-mixing makes it one of the most complex Arabic speech recognition challenges.
- Natural Arabic-French code-mixing within the same sentence
- French phonological and lexical influence
- English vocabulary in technical and professional contexts
- Regional variations: Beirut, North, South, Bekaa Valley
Native speakers
5M+
Lebanon
Diaspora
Worldwide
Latin America, Africa, Europe, Australia
Code-mixing
Trilingual
Arabic + French + English
WER
<16%
Hamsa benchmark 2024
The problem
Why generic Arabic STT fails
on Lebanese Arabic
A model trained on MSA achieves 38–54% word error rate on Lebanese speech. Here is why.
The solution
How Hamsa handles
Lebanese Arabic natively
Capabilities
Everything you need for
Lebanese Arabic STT
Transcript examples
Real Lebanese Arabic.
Accurate transcription.
Use cases
Where Lebanese Arabic STT
delivers results
Accuracy benchmark
Lebanese Arabic Speech Recognition Benchmark
Tested on real Lebanese speech across diaspora services, media, and financial environments. Result: under 16% WER compared to 38–54% for generic models.
Word Error Rate (WER) · lower is better
Measured on real Lebanese speech benchmarks · 2024 · Lower is better
Architecture
How it works
Developer API
Two lines to Lebanese Arabic speech recognition
Drop in the Hamsa SDK, set dialect: "lebanese", and start streaming transcripts. No preprocessing pipeline, no normalization to MSA.
import { Hamsa } from '@hamsa-ai/sdk';
const stream = hamsa.stt.stream({
model: 'hamsa-stt-s4',
dialect: 'lebanese',
diarization: true,
});
stream.on('transcript', (t) => console.log(t.text));
// → "كيفك؟ شو في جديد؟" · 00:01.420Related services
More Hamsa for Lebanese Arabic
Why Hamsa
Built for Arabic from the ground up
FAQ
Lebanese Arabic STT ·
frequently asked
Does the model understand Arabic-French code-mixing?
Yes. This pattern is fundamental in training data, not an exception.
Does it work with Lebanese diaspora speech?
Yes. Training data includes Lebanese speech from multiple diaspora environments.
Does it cover North and South Lebanon varieties?
Yes. Data from multiple Lebanese regions covers regional diversity.
Is it suitable for diaspora remittance services?
Yes. Transcription accuracy is appropriate for financial service environments.
Can custom financial vocabulary be added?
Yes, via a custom word list for product names and terminology.
Explore more



