Studio/For developers and businesses

A voice API that behaves like real infrastructure.

Koythu ships with the primitives a production system expects - predictable latency, streaming output, scope bound tokens, timing metadata, and dedicated capacity at business scale. No surprises the first time you plug it into an IVR.

koythu.render

import { Koythu } from '@koythu/sdk';

const koythu = new Koythu({ token: process.env.KOYTHU_TOKEN });

const audio = await koythu.render({
  voice: 'aanya',
  tone:  'warm',
  language: 'en-IN',
  script: 'Welcome. Your table is ready.',
  format: 'wav',
  stream: true,
});

await audio.pipe(fs.createWriteStream('welcome.wav'));

What the API gives you

Six primitives cover most of what a product needs.

REST API

A single endpoint per action - render, clone, list voices, list languages. Deterministic responses, well shaped errors.

Streaming audio

For live IVR and voice bots. Open a socket, stream text tokens, receive audio chunks with low first byte latency.

Predictable latency

Median first byte on standard voices stays under one second. Business tier gets reserved capacity for peak traffic.

Timing metadata

Phoneme, word and sentence boundaries returned with every render. Fits caption pipelines and IVR menu logic.

Scope bound tokens

Every API key is scoped to voices, languages, and usage caps. Rotate without downtime, audit without effort.

SDKs and examples

Reference implementations for Node, Python, Go and PHP. Copy paste starts working before you finish the API tour.

Where it fits

Three surfaces where a voice API changes the product.

IVR

Call flows that sound human

Replace legacy IVR prompts one call flow at a time. The same voice reads menus, greets returning callers, and confirms transactions.

Product voice

In app narration and confirmations

The onboarding tour, the delivery ETA, the confirmation of an action - all in a voice that matches the brand, in every language the product ships in.

Voice bots

Agents that do not stall

Streaming render pairs with your LLM tokens for a voice agent that speaks as it thinks. No pre buffer stall, no read out lag.

Connect with us about business tier See pricing