Questions

Frequently asked questions

Everything about the private beta, supported languages, and how the invite actually works.

Is my data private?

Yes. Recordings and transcripts stay on your iPhone. Nothing is uploaded to Nasar Flow's servers or anyone else's. See the full privacy policy for what the website itself collects (just your email, to send the invite).

Which languages are supported?

Singlish and English, Malay, and colloquial Arabic (beta), plus a general multilingual fallback. You can switch between them mid-sentence.

Do I need an internet connection?

No. Transcription runs fully on-device once you've downloaded a language model (roughly 190MB each for Singlish, Malay and Arabic, downloaded over Wi-Fi or cellular the first time).

Can I use it in WhatsApp, Messages or any other app?

Yes. Nasar Flow comes as a keyboard. Add it once (Settings → General → Keyboard → Keyboards → Nasar Flow, then allow Full Access), turn Flow on, and the keyboard's mic works everywhere you type. Full Access is needed only so the keyboard can hand off to the app: iOS doesn't let any keyboard use the microphone, so the app does the listening.

Can I say things like "new line" or "delete that"?

Yes, but only when you hold the mic instead of tapping it, so the words "new line" said normally stay words. Twelve commands in English, Malay and Arabic, plus "change X to Y" and "spell A-L-Y". The full list is in the app's README on GitHub.

Will it "correct" my Singlish into proper English?

No. Cleanup removes "um" and "uh", fixes capitals and spacing and, if you choose Full, writes numbers and prices the way you'd type them. It never strips lah, leh or lor, never translates, and the raw transcript is always kept.

Is it free?

Free during the beta.

What happens after I enter my email?

You're added straight to the TestFlight beta group via Apple's API. Apple emails you an invite link directly. Install the free TestFlight app first if you don't already have it.

How long does the invite take?

Usually a few minutes. If the beta is mid-review with Apple, it can take up to 48 hours the first time around.

I didn't get an email. What now?

Check spam first, and make sure you used the same email you signed up with. If it's been more than 48 hours, contact us or open an issue on GitHub.

Why is the Arabic model marked "beta"?

It's a colloquial (Amiyah) Arabic model, not Modern Standard Arabic, and it's genuinely less accurate than the Singlish and Malay models today. See the roadmap for what's being improved.

Hasn't code-switching speech recognition been done before?

Yes, and that's worth being upfront about. Researchers at AI Singapore's Speech Lab and commercial APIs like Soniox have been working on Singlish and Malay-English mixing for years. What none of them do: fold in the Arabic and Islamic phrases that show up mid-sentence in how Malay-Muslim communities actually talk, run entirely on-device instead of a cloud API, or ship as something you just download instead of an enterprise tool. That's the specific, narrower gap Nasar Flow is for.

Is the code open source?

Yes. Nasar Flow is built on whisper.cpp and the full source is on GitHub under the MIT license. See the credits page for the models and libraries it stands on.

What iPhone do I need?

iOS 17.0 or later, and roughly 200MB of free space per language model you download.