Arabic transcription
Updated
Use BroadNet Transcribe Arabic (broadnet/cohere-transcribe-arabic) for Arabic audio, at $0.0042 per minute. BroadNet Transcribe (broadnet/cohere-transcribe) covers 14 languages, Arabic included, at $0.002. Both take WAV or MP3 up to 26,214,400 bytes on /v1/audio/transcriptions. Send language=ar.

Arabic model or multilingual model?
| Model | Languages | Per minute | Per hour | Starting credit |
|---|---|---|---|---|
BroadNet Transcribe Arabicbroadnet/cohere-transcribe-arabic | Arabic | $0.0042 | $0.252 | Yes, about 1,190 minutes |
BroadNet Transcribebroadnet/cohere-transcribe | 14 languages, Arabic included | $0.002 | $0.12 | Yes, about 2,500 minutes |
BroadNet Transcribe Arabic
BroadNet's Arabic model. BroadNet says it handles Gulf, Egyptian, Levantine, Maghrebi and Modern Standard Arabic, and recommends it for Arabic audio. It returns Arabic script. Model page and BroadNet's reported results.
BroadNet Transcribe
One model for 14 languages, at a lower rate per minute. Use it when your audio mixes languages across files, and test its Arabic output on your own recordings first. Model page.
Which languages and formats are supported?
- 01Voice notePhones and messaging apps often save .ogg or .m4a
.ogg.m4a - 02ConvertWAV or MP3, up to 26,214,400 bytes per file
.wav.mp3ffmpeg -i voice-note.ogg -ar 16000 -ac 1 voice-note.wav - 03TranscribeArabic model, language set to Arabic
POST /v1/audio/transcriptionsmodel=broadnet/cohere-transcribe-arabiclanguage=ar - 04TextRead the text field of the response
text
Phone voice notes usually arrive as .ogg or .m4a. Convert them to WAV or MP3 of up to 26,214,400 bytes, send the file to the transcription endpoint with language set to ar, and read the text from the response.
- Endpoint
POST /v1/audio/transcriptions(multipart form)- Formats
- WAV (
audio/wav) or MP3 (audio/mpeg) - File size
- Up to 26,214,400 bytes (25 MiB) per file. A larger file returns 413; split or compress it.
- Language field
language=arfor Arabic. Always send it; without it, Arabic can come back in Latin letters.- Response
- JSON with the transcript in
text - BroadNet Transcribe languages
- English, Spanish, French, German, Italian, Portuguese, Greek, Dutch, Polish, Mandarin Chinese, Japanese, Korean, Vietnamese, Arabic
WhatsApp .ogg and phone .m4a files need converting first: ffmpeg -i voice-note.ogg -ar 16000 -ac 1 voice-note.wav
What will my audio cost?
Cost is audio minutes × the model's per-minute rate. Enter your monthly minutes to compare both models.
How do I integrate it?
Use the OpenAI SDK's transcription call or a plain multipart request. Keep the key in BROADNET_API_KEY on your server.
transcribe.py
import os
from openai import OpenAI
client = OpenAI(
base_url="https://inference.broadnet.ai/v1",
api_key=os.environ["BROADNET_API_KEY"],
)
with open("recording.wav", "rb") as audio:
result = client.audio.transcriptions.create(
model="broadnet/cohere-transcribe-arabic",
file=audio,
language="ar",
)
print(result.text)transcribe.sh
curl https://inference.broadnet.ai/v1/audio/transcriptions \
-H "Authorization: Bearer $BROADNET_API_KEY" \
-F model=broadnet/cohere-transcribe-arabic \
-F language=ar \
-F file=@voice-note.wavIllustrative example
{"text": "مرحباً، لقد وصلت إلى البريد الصوتي، اترك رسالتك من فضلك"}The response shape from BroadNet's reference, with a synthetic voicemail. Not a recorded result or an accuracy benchmark.
Next: create a key on inference.broadnet.ai · all quickstarts
Guides and articles on Arabic speech
- Arabic speech-to-text API, step by step
- Transcribe WhatsApp voice notes to text
- Transcribe WhatsApp voice notes in n8n
- Arabic speech-to-text API: how to choose one
- Arabic dialect speech recognition
- Arabic ASR word error rate explained
- Multilingual speech-to-text API
- Transcribe call center recordings
- Speech-to-text API pricing compared
- Speech-to-text cost calculator
- Speech-to-text API price index
Questions about Arabic transcription
Which model should I use for Arabic audio?
Start with BroadNet Transcribe Arabic (broadnet/cohere-transcribe-arabic), BroadNet's Arabic model, at $0.0042 a minute. BroadNet Transcribe also accepts Arabic among 14 languages at $0.002 a minute. BroadNet recommends the Arabic model for Arabic audio; test both on your own recordings.
Do I have to send the language?
Yes. Send language=ar with Arabic audio. Without it, Arabic can come back transliterated into Latin letters.
Which audio formats and sizes are accepted?
WAV (audio/wav) or MP3 (audio/mpeg), up to 26,214,400 bytes (25 MiB) per file. Convert WhatsApp .ogg or phone .m4a recordings to WAV with ffmpeg first.
Can I use the starting credit for transcription?
Yes. The $5 starting credit for new accounts covers both transcription models.
Is my audio stored?
Inference content is not retained. Necessary account, billing and security metadata is retained as described in BroadNet's inference privacy statement. Its list of content that is not kept includes audio and transcripts.