whatmodel.sh / tts / GPT-SoVITS v2

GPT-SoVITS v2

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

GPT-SoVITS v2 is an open-weights text-to-speech model from RVC-Boss, released August 2024. Its published size is 0.2B parameters. Licence: mit.

Catalogue identity notes

lj1995/GPT-SoVITS gsv-v2final-pretrained (s2G2333k.pth + s1bert25hz-5kh-longer-epoch=12-step=369668.ckpt), zero-shot (no fine-tuning). Distinct from v1, v3, v4 and v2Pro.

Model information sources

· Parameter count ↗

Official GPT-SoVITS repo wiki page "GPT-SoVITS-v2-features", linked from the official README V2 Release Notes ("more details"); every revision of the page was committed by RVC-Boss (the maker), last on 2024-08-07. v2 parameter count is stated as unchanged from v1 200M; rounded headline figure for the GPT+SoVITS system, the table does not split it by component.

44.0±3.5#80 of 118 text-to-speech modelsfrom 3 benchmark results across 2 sources

Score within the Vapi Humanness · Bradley–Terry rating, 2026-09-25 snapshot comparison set only.

$ how it compares

The text-to-speech models ranked either side of GPT-SoVITS v2, out of 148 with a score.

#modelscorewhereRAMreleased
100Amazon Polly Neural Amazon Web Services45±4api $16.00/M chars
101KittenTTS Nano 0.8 KittenML44±5comparison only2026-02
102GPT-SoVITS v2 RVC-Boss44±4local~1.1GB2024-08
103MeloTTS English MyShell44±5local~0.7GB2024-02
104Parler-TTS Mini v1 Hugging Face (Parler-TTS)44±6local~3.5GB2024-08

see all text-to-speech models · find the ones that fit your machine

$ what it's good at

other tests3 not in a task · 95% of its score
testresultvs the others in its rundifficultycountswhy
TTSDS2 listening test · speaker-similarity MOS (4 English domains)3.82 / 5 ✎▅·█▅▅·█▃▃▅·▅ beats 13/17easy · avg 4.03 / 5▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 82%at its level
TTSDS2 listening test · naturalness MOS (4 English domains)2.81 / 5 ✎▅··▃▅··▅█▅██ beats 5/17medium · avg 3.30 / 5▇▇▇ 8.0%blunt · at its level
TTSDS2 multilingual benchmark · English overall (YouTube, 2025 round)87.2% ✎▂▂█·▂·█▆▂·▄▂ beats 13/17very easy · avg 89%▇▇ 5.4%blunt

$ where to run it

Other ways to run it

version

  1. Run it as published · needs ~1.1 GB · from github.com/RVC-Boss (official)
    git clone https://github.com/RVC-Boss/GPT-SoVITS && cd GPT-SoVITS && conda create -n GPTSoVits python=3.10 && conda activate GPTSoVits && bash install.sh --device CPU --source HF && python api_v2.py -a 127.0.0.1 -p 9880 -c GPT_SoVITS/configs/tts_infer.yaml

Weights and documentation: huggingface.co/lj1995 ↗

$ how the score was worked out · TTSDS2 listening test results… 90%, TTSDS2 multilingual benchmark… 5%

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

sourceresultsshare of the scorelicence
TTSDS2 listening test results (ttsds/listening_test)2▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 90%MIT
TTSDS2 multilingual benchmark results (ttsds/leaderboard_data)1▇▇ 5%MIT
prior · pull toward the middle▇ 5%
weighted average= 44.0

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json