Emirati stays Emirati, and Egyptian stays Egyptian. We do not convert dialects to Fusha or guess what was said.
Record up to 60 seconds from your microphone, or drop an audio file to transcribe.
Arabic covers every dialect and mixed speech. Pick English only for speech that is all English.
What comes back
This is not Fusha with a dialect added on. Each dialect has its own training track, so the words people actually use appear as spoken, without errors.
Most Arabic transcription tools give you a draft that needs correcting. Here, you get the conversation as it happened: who said what, when, and the exact words they used.
مب stays مب. النهارده stays النهارده . The word the speaker chose is the same word in the file.
Caller and agent, host and guest, two people in a meeting. Each turn carries its speaker, even when they overlap.
Jump to the audio position from the text. Export as plain text, subtitles or JSON with a timestamp for every word.
Search yesterday's calls by what was said. Review the quality of every call and give Sada what people actually ask about.
Interviews, series, podcasts. SRT out of the box, timed to the word, in the dialect the guest spoke.
The WhatsApp voice note, the site visit, the majlis meeting. Record it, read it, share it.
In the street, car or open office, run the file through Sound Cleaning before transcription for clearer, more accurate words. The same account, the same file.
Send the file and receive the transcript within the same API request. The system automatically detects the dialect from the audio, so there are no settings to configure. Longer files are processed as a separate job.
# python import requests API = "https://api.echonova.ai" AUTH = dict(Authorization="Bearer YOUR_API_KEY") with open("call.mp3", "rb") as audio: res = requests.post( API + "/v1/speech-to-text/sync", headers=AUTH, files=dict(file=("call.mp3", audio, "audio/mpeg")), ) print(res.json()["data"]["transcription"]) # → هلا، أنا اتصلت أمس بخصوص الفاتورة
Security & sovereignty
Speech to TextConvert audio to text in real time or in batches, across eight dialects.
Text to SpeechNatural Arabic voices in the caller’s own dialect.
Noise and echo are removed before the model listens to the audio.
From calls, apps or recordings.
It turns spoken Arabic in an audio recording into written text. Try the demo with your own recording to assess the result for your use case.
Start with the speech-to-text API documentation for authentication, requests and the available response formats.
Sovereign Arabic voice AI
