본문으로 건너뛰기

Kits AI

실시간 음성 복제

별도의 학습 없이도 더 뛰어난 품질의 음성 복제를 지원하는 새로운 모델입니다.

실시간 음성 복제

April 23, 2025

by Anastasiia Herus

We’re excited to introduce Instant Voice Cloning (IVC), the first product powered by zero-shot singing voice conversion, now available at kits.ai. IVC delivers unparalleled singer similarity and quality with just 15-30 seconds of reference audio and no training time, making it easier than ever to explore new vocal ideas instantly.

Input

Target Singer Reference

Output

A major step forward in singer similarity

Instant Voice Cloning 아키텍처: 입력 오디오를 피치·콘텐츠·스타일 인코딩으로 나누고(스타일은 레퍼런스 오디오에서 추출), 디퓨전 트랜스포머, 보코더, 후처리를 거칩니다

One of the most common challenges with zero-shot voice conversion is achieving a compelling level of similarity to the reference singer without sacrificing overall quality. Thanks to a trio of key innovations, IVC pushes the boundaries of likeness alongside strides in fidelity and realism.

A larger and more curated dataset of studio-recorded, licensed vocals. We’ve expanded and cleaned up the dataset to improve performance across a wider range of vocal styles and timbres.

A diffusion transformer module is used to achieve greater fidelity during synthesis, especially when reproducing the expressiveness of the reference voice.

A new, singing-specific vocoder designed and trained to reproduce singing characteristics with greater nuance and realism.

Our previous research in zero-shot voice conversion provides the foundation for these improvements. Instant Voice Cloning utilizes the same three-part feature embedding approach to disentangle melody, content, and style.

This architecture makes it possible to transform the vocal identity and expressiveness of an input file with impressive accuracy, all with only seconds of reference audio.

Trained on high-quality licensed data

As always, we prioritize quality and artist respect in everything we build. The Instant Voice Cloning system was trained on a dataset of studio-recorded vocals, licensed directly from artists. Every vocal stem was hand-processed by professional audio engineers to meet release-level quality standards. This ensures our models are not just technically powerful — they’re artistically trustworthy.

Try it now

IVC is available for free to all users for a limited time — try it now.

Kits로 창작을 시작하세요

음성 변환, 음성 제작 및 오디오 도구를 살펴보세요. 앱에서 현재 요금제와 계정 조건을 확인하세요.

무료로 시작하기