Today, we are excited to release KiriTTS — our new Text-to-Speech model designed for real-world, production-ready applications.

We built KiriTTS because we were not satisfied with what was available. Existing TTS solutions were either too slow for live applications, too robotic for natural-sounding speech, or had little to no support for Southeast Asian languages. KiriTTS is our answer to all three problems.

What KiriTTS can do

Multilingual speech generation — KiriTTS supports a wide range of languages out of the box, including Khmer, English, and more. No language tag needed; just pass in your text and the model handles the rest.

Voice cloning — Provide a short audio clip of any speaker and KiriTTS will clone that voice faithfully. It preserves tone, rhythm, and character across long-form outputs without drifting.

Voice design — Do not have a reference clip? Describe the voice you want in plain language — age, gender, tone, pace — and KiriTTS generates a brand new voice from scratch.

Real-time streaming — KiriTTS is built for low-latency applications. Audio starts playing in milliseconds, making it suitable for voice assistants, live narration, and interactive agents.

Studio-quality output — Every audio output is clean, natural, and consistent — ready to use in production without additional post-processing.

Who it is for

KiriTTS is designed for developers and teams building:

Voice assistants and conversational AI
Audiobook and podcast narration tools
Accessibility features for visually impaired users
E-learning and educational platforms
Customer service automation

Why Khmer matters to us

We are building KiriTTS from Cambodia, for the world. Khmer is notoriously underserved by mainstream TTS providers. Most models mispronounce it, skip it entirely, or produce robotic output that native speakers immediately reject.

KiriTTS takes Khmer seriously. We trained on real Khmer speech data and made sure the model handles the language's complex script and unique phonetics correctly.

Get started

KiriTTS is available now via API. The fastest way to integrate is through our OpenAI SDK-compatible endpoint — if you are already using OpenAI's TTS, switching to KiriTTS takes just two lines of code.

JavaScript

import OpenAI from 'openai'

const client = new OpenAI({
  apiKey: 'YOUR_API_KEY',
  baseURL: 'https://api.kiritts.com/v1',
})

const response = await client.audio.speech.create({
  model: 'kiritts',
  voice: 'kiri',
  input: 'Hello, world!',
})

Python

from openai import OpenAI

client = OpenAI(
    api_key="YOUR_API_KEY",
    base_url="https://www.kiritts.com/api"
)

response = client.audio.speech.create(
    model="kiritts",
    voice="kiri",
    input="Hello, world!"
)
response.stream_to_file("output.mp3")

Point baseURL to https://www.kiritts.com/api, swap in your KiriTTS API key (which you can get by signing up at kiritts.com), and you are done. No SDK changes, no new dependencies.

Full API reference, available voices, parameters, and pricing are all in the docs:

→ kiritts.com/docs

What is next

This is just the beginning. We are actively working on expanding language coverage, improving emotional expressiveness, and reducing latency further for on-device use.

If you build something with KiriTTS, we would love to hear about it. Reach out directly or join our community Discord.

KiriTTS · Built in Cambodia, for the world.