Private beta · iOS 17+

Built for how we actually talk.

Nasar Flow transcribes Singlish, Malay, Arabic and English, the way people actually speak across Singapore, Malaysia, and Indonesia, even when you switch between them mid-sentence. It runs entirely on your iPhone. Nothing you say is ever uploaded.

Free during the beta. One email, no spam, just your TestFlight invite.

Made for sentences like these

Real vocabulary Nasar Flow is tuned to catch, not a transcript of a specific recording.

“Bismillah, let's go makan lah”

Arabic + Malay + Singlish

“Wah shiok sia, we go makan then balik house”

Singlish + Malay

How it works

Three steps. Zero uploads.

01

Speak naturally

Talk the way you already do, switching between Singlish, Malay, Arabic and English mid-sentence, no need to pick one.

02

Transcribed on your iPhone

Nasar Flow runs whisper.cpp entirely on-device. Your voice never leaves your phone. No internet required once a model is downloaded.

03

Read, copy, export

Every transcript stays on your device, ready to copy, search back through, or export whenever you need it.

Four models, one app

Tuned for how you actually code-switch

Nasar Flow auto-detects which language you're in and routes to the right model, including the everyday Arabic and Islamic phrases that slip into conversation.

Singlish & English

The broadest net: this is the default model, tuned on real Singlish speech. Shares enough slang with Malaysian English to hold up well there too.

lahlehlor shiokwalaoalamaksteady

Malay

Everyday Bahasa, not just formal Malay. The core vocabulary carries across Malaysia and Indonesia too.

makanjalanbalik tolongterima kasihselamat

Arabic Beta

Colloquial (Amiyah) Arabic, not Modern Standard, including the Islamic phrases that show up mid-sentence. Actively improving; expect rough edges during beta.

BismillahAlhamdulillahInsyaAllahAssalamualaikum

English / Multilingual

A general-purpose fallback model for everything else, straight from whisper.cpp.

General fallback

Private by design

Your voice never leaves your phone.

No servers, no cloud transcription, no account. Nasar Flow downloads a speech model once, then runs entirely offline.

  • No account required
  • No audio ever uploaded
  • Works fully offline
  • Open source on GitHub

Built in the open

Why I'm building this

Singlish, Malay, and Arabic code-switching (how Malay-Muslim communities across Singapore, Malaysia, and Indonesia actually speak day to day) has never been a first-class citizen in speech-to-text. Most transcription tools don't know Singlish, don't know Malay, and fall over the moment a "InsyaAllah" or "Bismillah" shows up mid-sentence.

Nasar Flow runs on whisper.cpp, entirely on-device, and I'm building the whole thing without a Mac, compiling and shipping every release through GitHub Actions. The source is public. It's early, and it'll have rough edges. That's exactly what the beta is for. See the credits page for the open models and libraries that make it possible.

Aarif, Nasar Pte Ltd

Request your invite

Get on the list.

Nasar Flow is invite-only on TestFlight. Enter your email and you're added automatically, no waiting on me. Apple usually sends the invite within a few minutes; if the beta is mid-review it can take up to 48 hours.

Requires iOS 17 or later. Free during the beta.