---
title: "k2-fsa's OmniVoice Clones Any Voice Across 600 Languages at 40x Real-Time Speed"
subtitle: "k2-fsa open-sources OmniVoice, a zero-shot TTS model covering 600+ languages at 40x real-time speed with Apache 2.0 licensing"
url: "https://alphasignal.ai/news/k2-fsa-s-omnivoice-clones-any-voice-across-600-languages-at-40x-real-time-speed"
author: "AlphaSignal Newsroom"
author_url: "https://alphasignal.ai/editorial-team"
company: "AlphaSignal"
company_url: "https://alphasignal.ai/company/alphasignal"
published: "2026-06-11T07:13:09.000Z"
updated: "2026-06-13T13:12:17.614Z"
type: "Repo"
topics:
  - "Audio"
  - "Open Source"
subtopics:
  - "Voice Cloning"
  - "Text To Speech"
upvotes: 36
source_url: "https://github.com/k2-fsa/OmniVoice"
comments: 0
truncated: true
---

# k2-fsa's OmniVoice Clones Any Voice Across 600 Languages at 40x Real-Time Speed

> k2-fsa open-sources OmniVoice, a zero-shot TTS model covering 600+ languages at 40x real-time speed with Apache 2.0 licensing

The k2-fsa team , the research group behind Kaldi and k2, with Daniel Povey as a core contributor , has open-sourced [OmniVoice](https://github.com/k2-fsa/OmniVoice), a zero-shot text-to-speech model that covers more than 600 languages. That is not a typo. Most production TTS systems top out at a few dozen languages. OmniVoice blows past that ceiling and does it without needing any fine-tuning on a target voice , just a short audio clip.

## The problem with existing multilingual TTS

Most competitive TTS models today , including XTTS v2 , support somewhere between 17 and 30 languages. Scaling further has historically meant either training separate models per language or accepting severe quality degradation on low-resource languages. The standard architecture compounds this: a two-stage pipeline that first converts text to semantic tokens, then converts those to audio, introduces error propagation at every step.

At its core, OmniVoice uses a novel diffusion language model-style discrete non-autoregressive (NAR) architecture. Unlike conventional discrete NAR models that suffer from performance bottlenecks in complex two-stage pipelines, OmniVoice directly maps text to multi-codebook acoustic tokens. Cutting out the intermediate semantic stage removes a major source of quality loss at scale.

## How it actually works

OmniVoice uses a diffusion language model architecture. Unlike traditional autoregressive TTS, this approach uses a Qwen LLM backbone to perform iterative unmasking/diffusion over multi-codebook discrete audio tokens. Think of it like BERT-style masked language modeling, but applied to audio tokens instead of text , the model learns to predict masked audio tokens from context, then iteratively refines the full sequence at inference time.

_Preview only. Full text requires an AlphaSignal Pro account._

## Sources

- [k2-fsa/OmniVoice — GitHub](https://github.com/k2-fsa/OmniVoice)
- [OmniVoice: Towards Omnilingual Zero-Shot Text-to-Speech with Diffusion Language Models — arXiv](https://arxiv.org/abs/2604.00688)
- [k2-fsa/OmniVoice — Hugging Face](https://huggingface.co/k2-fsa/OmniVoice)
- [OmniVoice: Open-Source TTS with 600+ Languages — DEV Community](https://dev.to/_46ea277e677b888e0cd13/omnivoice-open-source-tts-with-600-languages-and-zero-shot-voice-cloning-1mpn)
- [OmniVoice Paper Page — Hugging Face](https://huggingface.co/papers/2604.00688)