Moon

A voice AI that speaks
any language, anywhere.

Moon is the realtime voice layer for modern products. Native accents in forty two languages, sub second first byte, and a single line to speak from anywhere on the planet.

Try Moon Listen to the voices →
▶

Realtime, everywhere.

First byte lands in under two hundred milliseconds, from any region to any listener. Predictable enough to build a call center on, warm enough to build a bedtime story on.

Forty two languages, native register.

Every voice ships with the accent, breath and cadence a native speaker would recognise. Switch language mid sentence, keep the personality.

An SDK your engineers
can ship on.

One import, one line to speak, one line to stream. Typed schemas, deterministic latency, and the specific control a production engineer expects.

1
Prompt arrives
Any language, any client, any device
2
Moon composes the phrase
Model streams tokens as it thinks
3
First audio byte, 142ms
Listener hears speech before it is finished
4
Full response streamed
Native accent, natural breath, on device or in the cloud
VoiceLanguageLatencyStatus
Sophie
Warm, close mic
US English142msStreaming
Sena
Whispered, editorial
Japanese148msStreaming
Marcus
Baritone, calm
UK English138msStreaming
Iris
Bright, playful
French156msStreaming
Tomas
Cinematic, deep
Spanish144msStreaming
app/speak.ts
// One import, one line to speak

import { Moon } from "@moon/voice"



const voice = new Moon({ cast: "sophie" })



for await (const chunk of voice.speak(text)) {

  audio.write(chunk)

}

// First byte in 142ms, streamed to the ear.

Voice primitives,
not models to babysit.

◐

Sub second first byte

Every phrase begins before the model finishes composing it. Latency you can build a phone call on, not just a podcast.

◑

Native accents

Forty two languages, one hundred voices, and the small half smile inside a word. Speakers your users mistake for people.

◒

Predictable, everywhere

On device, in your cloud, or on ours. Deterministic latency, typed schemas, and the specific control production teams need.

Moon is the first voice API we shipped in production
without asking a user to forgive the machine.

Priya RamHead of Product, Northline
Northwind
Constellation
Halcyon
Bright Union
Field Guide
Longtable

Ship a voice
this afternoon.

Free on the first hundred thousand seconds. One import, one line to speak, and the first byte arrives inside an hour.

Try Moon Listen to the voices