Choosing a Model

All MenaVoice text to speech models cost the same, so choose on sound and behavior alone. Start from what your script needs.

​Quick guide

​What to consider

​Consistency

MenaVoice 1v is built for consistency: regenerated audio sounds much like the last take, which suits audio you update often, like notifications or product copy. MenaVoice 2 varies noticeably between takes. That's part of what makes it expressive, but it means you should generate a few takes, keep the one you like, and store it instead of generating it again.

​Speed

MenaVoice 1v is built for speed. When a person is waiting for the audio, such as in an app or on a phone line, it's the safest choice.

​Dialects

MenaVoice 2, 1.5 and 1v all support dialectId. MenaVoice Clone doesn't: a cloned voice keeps the accent of the recording it was made from, so clone a speaker of the dialect you want.

​Voices

The 30 studio voices work with MenaVoice 2, 1.5 and 1v. MenaVoice Clone works only with voices in your library.

​Compare them on your own script

The best test is your real script. Generate it with each model and listen side by side:

import base64
import os

import requests

script = "اكتشف عروض نهاية الأسبوع، خصومات تصل إلى خمسين بالمئة على كل المنتجات."

for quality in ["flash", "pro", "v2"]:
    response = requests.post(
        "https://api.menavoice.ai/api/tts",
        headers={"x-api-key": os.environ["MENAVOICE_API_KEY"]},
        json={"text": script, "voiceId": "sara", "quality": quality},
        timeout=120,
    )
    response.raise_for_status()
    with open(f"sample-{quality}.mp3", "wb") as f:
        f.write(base64.b64decode(response.json()["audio"]))

You can also compare models without code in the web app's Text to Speech page, where the Takes panel keeps each result for comparison, or in the API Playground.