GPT-SoVITS

GPT-SoVITS is an open-source voice cloning and text-to-speech toolkit with more than 63,000 stars on GitHub that can clone a voice from as little as one minute of reference audio.

Overview

GPT-SoVITS combines a GPT-style language model with the SoVITS (SoftVC VITS) singing/speech conversion lineage to deliver few-shot voice cloning: give it a short reference recording and it synthesizes new speech in that voice, with zero-shot inference from a five-second sample and fine-tuning from as little as one minute of data. The toolkit is cross-lingual, supporting inference in languages different from the training data — English, Chinese, Japanese, Korean, Cantonese and more — and it ships the full production chain: voice separation, automatic training-set segmentation, Chinese/Japanese annotation tools, a WebUI for training and inference, and a fast inference API. It has trended 45 times in the archive, is MIT-licensed, and was actively developed in October 2026.

GPT-SoVITS grew out of the RVC voice-conversion community and became the default open-source answer for anime, game and VTuber voice work as well as practical dubbing. Its training WebUI is designed for non-engineers: upload recordings, let the pipeline slice and label them, train, and synthesize — a workflow that previously required stitching together several research codebases by hand.

Key Features

• Few-shot cloning – fine-tune a convincing voice model from one minute of audio.
• Zero-shot TTS – clone and speak from a five-second reference with no training.
• Cross-lingual synthesis – generate languages beyond the training data’s language.
• Complete toolchain – separation, segmentation, annotation and training in one WebUI.
• Fast inference API – serve cloned voices to your own applications.
• Community ecosystem – extensive guides and shared models around the project.

💳 Pricing

GPT-SoVITS is free and open source under the MIT license. You run it on your own GPU; there is no subscription or per-character fee.

Who Is It For

Creators, VTubers, game developers, dubbing hobbyists and researchers who need custom cloned voices locally — with the responsibility to only clone voices they have the right to use.