# How to transcribe Arabic voice notes (WhatsApp, Telegram and phone recordings)

> In the Gulf, a lot of business happens in voice notes — orders, complaints, instructions. Here is how to turn them into text you can search, forward and act on, in the dialect they were spoken in.

- Source: https://nutq.dev/blog/transcribe-arabic-voice-notes
- Published: 2026-09-27
- Updated: 2026-09-27
- Author: nutq team (https://nutq.dev)

## Short answer

To transcribe an Arabic voice note, upload the file — a WhatsApp .opus, an .m4a or an .amr call recording — to a transcription tool, or POST it to a speech-to-text API. nutq takes these formats as they are, keeps Gulf and Levantine dialect as spoken, and costs $0.15 per hour of audio, so a one-minute voice note costs a fraction of a cent. New accounts get $5 of free credit, about 33 hours.

Across the Gulf, a voice note is how people actually communicate: a customer describing a problem, a supplier confirming an order, a manager giving instructions from the car. They are quick to send and slow to deal with — you cannot search them, skim them or forward them to someone who is in a meeting. This guide shows how to **transcribe Arabic voice notes** into text, in the dialect they were spoken in.

## Why are Arabic voice notes hard to transcribe?

Voice notes combine every difficulty at once:

- **Dialect.** Nobody records a voice note in Modern Standard Arabic. A model that rewrites dialect into MSA gives you a tidy text that is not what was said — see [Gulf Arabic vs MSA](/blog/gulf-arabic-vs-msa-transcription).
- **Compressed audio.** WhatsApp and similar apps compress voice notes heavily to save bandwidth.
- **Real-world noise.** Cars, cafés, children, air conditioning, a phone held at arm's length.
- **Mixed language.** English product names and numbers inside Arabic sentences.

## How do you transcribe a voice note without code?

On the [Arabic transcription](/arabic-transcription) page:

1. **Drop the file.** Save the voice note from WhatsApp or Telegram and drop it in. It stays in your browser until you sign in.
2. **Create a free account.** $5 of credit lands in your account, with no card.
3. **Read the transcript.** Copy it, download it, or find it later in your history.

Formats go in as they are — no converting first:

| Source | Typical format |
|---|---|
| WhatsApp voice note | `.opus` |
| Telegram voice message | `.ogg` |
| iPhone voice memo | `.m4a` |
| Phone call recorder | `.amr`, `.m4a` |
| Video from a phone | `.mp4`, `.mov` |

## How do you transcribe voice notes with the API?

For volume — a support inbox, a sales team, a CRM — send each file to the API:

```bash
curl -X POST https://nutq.dev/v1/transcribe \
  -H "Authorization: Bearer $NUTQ_API_KEY" \
  -F "file=@voice-note.opus"
```

The response carries the text, the detected language and exactly what the request cost:

```json
{
  "text": "هلا، أبا أغيّر موعد التوصيل لباكر إذا ممكن",
  "language": "ar",
  "usage": { "audio_seconds": 6.4, "cost_usd": 0.00027 }
}
```

The same request works from an automation tool: an HTTP Request node in n8n that uploads the file as form data, then passes the text to the next step. Our [n8n text-to-speech guide](/blog/arabic-text-to-speech-n8n) covers the credential setup, which is identical. The full request is in the [Arabic speech to text API guide](/blog/arabic-speech-to-text-api).

## Which dialects and languages work?

| Speech | Result |
|---|---|
| Gulf Arabic (Emirati, Saudi, Kuwaiti, Qatari…) | Kept as spoken |
| Levantine Arabic | Kept as spoken |
| Modern Standard Arabic | Transcribed as written Arabic |
| Maghrebi, Yemeni, Sudanese Arabic | Recognised, less accurately |
| 13 other languages, including English and French | Detected per file |

Leave the language on auto and each voice note is detected on its own, so an inbox with a mix of Arabic and English messages needs no sorting first.

## What does it cost?

| Voice note length | Cost at $0.15 per hour |
|---|---|
| 30 seconds | $0.00125 |
| 1 minute | $0.0025 |
| 5 minutes | $0.0125 |
| 1 hour of voice notes | $0.15 |

Billed by the second, and failed requests are never billed. The $5 of free credit covers about 33 hours of audio.

## What happens to the recordings?

The recording goes to the transcription server, is transcribed and is discarded. Only the text is kept in your history, and you can delete it at any time. If voice notes cannot leave your organisation at all, the same model runs [on-premise inside your own network](/blog/on-premise-arabic-speech-recognition).

## Tips for better transcripts

- **Send the original file.** Re-recording or converting a voice note loses quality.
- **Leave the language on auto** unless you know it; the response says what was detected.
- **Check names and numbers.** They are the words a reader acts on, so skim them first.
- **Keep the transcript with the audio.** Store the text next to the original file, so anyone who doubts a word can play that moment back.

Try it with your own voice notes at [sign-up](/sign-up) — the first $5 is free.

## FAQ

### Can I transcribe WhatsApp voice notes in Arabic?

Yes. WhatsApp voice notes (.opus) go in as they are, with no conversion. The transcript keeps Gulf and Levantine dialect as spoken.

### Which audio formats work?

Any common audio or video format, including OPUS, OGG, M4A, AMR, MP3, WAV, AAC, FLAC, MP4 and MOV, up to 100 MB per file.

### How much does it cost to transcribe a voice note?

$0.15 per hour of audio, billed by the second. A one-minute voice note costs a quarter of a cent. New accounts get $5 of free credit with no card.

### What happens to the recording?

It is transcribed and discarded. Only the text is kept, in your history, until you delete it.

### Can I automate voice note transcription?

Yes. POST each file to /v1/transcribe from your own code or from an automation tool such as n8n, and route the text wherever you need it.