How far? Pidgin

Voice AI that sounds like us.

A realtime voice API for Nigerian English, Pidgin, Yorùbá, Igbo and Hausa, mixed the way people really talk.

Listen in

A conversation, not a form.

Today's voice AI sounds British or American, and it stops listening the moment you switch languages. Oya keeps up: you can mix, interrupt and think out loud, and it answers the way a friend on the phone would.

9:41

OYABoma · Auto

Oya is speaking

Abeg, how I go take reach Ibadan from Lagos this weekend?

Ehen! If na weekend, comot early, like six for morning, before traffic hold for Ojota.

2.9 s · median 3.0 s

00:14

Mix freely

Pidgin, English, both at once.No language menu to set. Oya hears the switch mid-sentence and answers in kind.

Sound local

Ehen, not "Certainly!"Replies are written for the ear, with real reactions, spoken numbers and local voices.

Interrupt

Talk over it, like a real conversation.The moment you speak, Oya stops and listens. No waiting for it to finish.

Languages · Live in Nigeria

Five languages. One conversation.

Every voice can speak every language, so a speaker in Lagos and one in Kano both hear someone who sounds like home.

PidginHow far? Wetin dey happen?Boma · Ufoma · Justice · Tega

EnglishGood morning, how are you?Remi · Lucy · Kingsley · Jude

YorùbáBáwo ni?Sade · Funmi · Femi · Segun

IgboKedụ?Ngozi · Amara · Obinna · Ebuka

HausaSannu!Amina · Zainab · Aliyu · Hasan

Coming soon · Ghana

TwiƐte sɛn?Soon

GaTe oyɔɔ tɛŋŋ?Soon

EweŊdi na woSoon

Pidgin & EnglishChale, how you dey?Soon

Then across Africa

Later

SwahiliHabari?East Africa
ArabicأهلاًNorth Africa
AmharicሰላምEthiopia
OromoAkkam?Ethiopia
ZuluSawubona!South Africa

For developers

One socket. Audio in, voice out.

Stream the speaker's audio over a WebSocket and get Oya's voice back as it's spoken, with transcripts and timings alongside.

Streaming both ways

PCM16 in at 16 kHz, voice out at 24 kHz, sentence by sentence.

Barge-in built in

Talk over a reply and it stops; stale audio is easy to drop.

Code-switching by default

Leave the language on auto, or pin one per session.

Connectws /v1/realtime

const oya = new WebSocket("wss://…/v1/realtime");
oya.binaryType = "arraybuffer";

oya.onopen = () => oya.send(JSON.stringify({
  type: "session.update",
  session: { language: "", voice: "boma" },  // "" = auto
}));

Quickstart, every event, the audio framing, the turn lifecycle, voices and honest limits.

Read the developer docs

Roadmap