Kikiri German — Martin Harbecke

German single-speaker TTS model fine-tuned on the Martin Harbecke voice using StyleTTS2 Stage 2.

Built on top of kikiri-german-base-51speakers-synthetic.

Demo

StyleTTS 2 — Abstract (German)

Model Details

Usage

# Uses the kokoro library as underlying framework
from kokoro import KModel, KPipeline
import torch

kmodel = KModel(repo_id="hexgrad/Kokoro-82M", config="config.json", model="kikiri_german_martin_ep10.pth").to("cuda").eval()
pipeline = KPipeline(lang_code="de", model=kmodel)
voice = torch.load("voices/martin.pt", map_location="cpu")

text = "Guten Tag, wie geht es Ihnen?"
for _, _, audio in pipeline(text, voice=voice):
    # audio is a numpy array at 24kHz
    pass

Files

License

Apache 2.0