forkalt.

MITfreeRuns locallyVideo & Audio

GPT-SoVITS: Open-Source Voice-Cloning TTS

MIT few-shot voice-cloning TTS. One minute of data is the project’s own claim.

GPT-SoVITS GitHub social card used to identify the few-shot TTS project
Official project image. Source · License: MIT.
License
MIT
Stars
63,196 (2026-10-19)
Platforms
Windows, Linux
First seen
2024-01-14

TL;DR. GPT-SoVITS is MIT few-shot voice-cloning TTS. One minute of data is the project’s own claim. License: MIT (verified). Platforms: Windows, Linux. Price: free. Runs locally: yes. It is an alternative to ElevenLabs.

NameGPT-SoVITS
LanguagePython
LicenseMIT
Pricingfree
PlatformsWindows, Linux
Runs locallyYes
Stats63,196 stars · 6,712 forks (as of 2026-10-19)
First seen2024-01-14
Repositoryhttps://github.com/RVC-Boss/GPT-SoVITS

GPT-SoVITS is MIT few-shot TTS. Official host we opened on 2026-10-19: RVC-Boss/GPT-SoVITS (63,196 stars, 6,712 forks, last push 2026-10-19).

It hangs off ElevenLabs alternatives next to Drama 3. Related: Descript alternatives (whisper.cpp).

FAQ

Is GPT-SoVITS an ElevenLabs alternative?

It is MIT local TTS with few-shot cloning (63,196 stars, pushed 2026-10-19). ElevenLabs is hosted. No quality ranking. See ElevenLabs alternatives.

How much audio do I need?

The GitHub description says one minute of voice data can train a good model. That is an upstream claim. Untested here.

How does it relate to Drama 3 / fish-speech?

Drama 3 is the open-weight Fish Audio checkpoint already on the ElevenLabs hub. GPT-SoVITS is a separate MIT training/inference tree.

Sources & verification

Last verified 2026-10-19 · Last updated 2026-10-19· Edited by forkalt editors · Editorial policy

License notes are not legal advice. Read the original LICENSE. Trademark & disclaimer.