opendunes

Nova-3

by DeepgramReleased 15 July 2026

Nova-3 takes an audio recording and gives back the written transcript. It copes with speech in noisy places such as call centres and drive-thrus, and it can follow a conversation that switches between languages, among them English, Spanish, French, German, Hindi, Russian, Portuguese, Japanese, Italian and Dutch. Use it to transcribe support calls, meetings and voice notes, live or from files. It only writes text; to speak text back, pair it with Aura-2 or Flux TTS.

TakesAudio
GivesText
Open in chatCreate API keyCompare

What it costs

You pay for what each request uses. No subscription, no minimum.

1,43 DA per minute of audio3 000 DA buys 2 000 minutes of audio

One minute of audio, the ordinary use

60 seconds of audio sent× 0,024 DA1,43 DA
Total1,43 DA

Every rate

Meter↑ you send↓ comes back
Second of audio0,024 DA

per second of audio sent

Transcribe a one-minute voice note≈ 1,43 DA
Caption a 5-minute live stream≈ 7,16 DA
Transcribe a 10-minute customer call≈ 14,31 DA
Transcribe a 30-minute podcast episode≈ 42,94 DA
Transcribe a one-hour meeting≈ 85,87 DA
When you pay
After the answer, from your balance. Nothing is held.
Why it varies
The length of the audio.
After each one
The exact charge shows on the request.
DepositDeposits from 50 DA, with Edahabia or CIB.

Where your request runs

One company runs this model. OpenDunes sends every request to it.

Deepgram…
Per second of audio
0,024 DA
Latency
…
ProviderPer second of audioLatencyUptime
Deepgram0,024 DA……

You pay the price of the provider that served your request. Prices read 15:18 · speed over the last 30 minutes · uptime over 3 days.

Call it in code

Swap the base URL and the key. Your OpenAI code runs unchanged.

POST /v1/audio/transcriptions
Base URL
https://opendunes.com/v1
Model ID
deepgram/nova-3
import os
from openai import OpenAI

client = OpenAI(
    base_url="https://opendunes.com/v1",
    api_key=os.environ["OPENDUNES_API_KEY"],
)

with open("meeting.mp3", "rb") as f:
    resp = client.audio.transcriptions.create(model="deepgram/nova-3", file=f)

print(resp.text)
pip install openaiFull reference for this route
Sign in to run it

Runs this request once on your balance.

Questions about Nova-3

  • How much does Nova-3 cost in dinars?

    About 1,43 DA per minute of audio. ↑ 0,024 DA per second of audio.

  • Is Nova-3 free? Can I try it?

    No. You pay per use from your balance, with no subscription. Try it in Chat at the same price.

  • What is Nova-3 best for?

    Nova-3 takes an audio recording and gives back the written transcript. It copes with speech in noisy places such as call centres and drive-thrus, and it can follow a conversation that switches between languages, among them English, Spanish, French, German, Hindi, Russian, Portuguese, Japanese, Italian and Dutch. Use it to transcribe support calls, meetings and voice notes, live or from files. It only writes text; to speak text back, pair it with Aura-2 or Flux TTS.

  • How do I pay for Nova-3 without an international card?

    Deposit dinars with your Edahabia or CIB card. Every model, Nova-3 included, is paid from that balance. No foreign card, no conversion.

  • Does Nova-3 work in Arabic, French or Darja?

    Deepgram lists many languages, among them French and Arabic with regional codes including ar-DZ for Algeria and ar-MA and ar-TN for Morocco and Tunisia. Darja is not named as such.

  • Does it say who spoke, and when?

    Speaker labels: not stated. Word timings: not stated.

  • How long can a file be?

    Up to 25 MB as an upload. Longer audio goes in the request body.

  • Who makes Nova-3, and when was it released?

    Deepgram. Released 15 July 2026.

Other Deepgram models, priced in dinars

See all Deepgram models
Aura-2Deepgram · Jul 16, 2026

Aura-2 turns text into speech in more than 40 voices from Deepgram, and it is the Deepgram choice when the language is not English: Spanish, German, French, Dutch, Italian and Japanese are covered. It starts speaking in under 200 milliseconds, which suits phone agents, automated menus and customer-service bots. For English-only builds with more expressive delivery, Flux TTS is the newer voice; Orpheus 3B fits English storytelling.

90 voices
≈5,99 DAper paragraph read aloud
View model
Flux TTSDeepgram · Aug 12, 2026

Flux TTS turns English text into speech built for live conversation, with 36 voices across seven accents. Delivery is expressive by default, so a voice agent sounds like it is replying, not reading. Use it for voice agents, phone support and general English narration in a new build.

36 voices
≈8,99 DAper paragraph read aloud
View model