Loading…
Loading…
hexgrad/Kokoro-82MKokoClone is a fast, real-time compatible multilingual voice cloning system built on top of Kokoro-ONNX, one of the fastest open-source neural TTS engines available today.
It allows you to:
Just text → voice → cloned output.
KokoClone is powered by Kokoro-ONNX, a highly optimized neural TTS engine designed for:
Unlike many heavy TTS systems, Kokoro is lightweight and responsive — making KokoClone suitable for real-time applications, voice assistants, demos, and interactive tools.
Generate native speech in:
en)hi)fr)ja)zh)it)pt)es)Upload a short voice sample and KokoClone transfers its vocal characteristics to the generated speech.
Built on Kokoro’s efficient ONNX runtime pipeline, KokoClone runs smoothly on:
On first run, required model files are downloaded automatically and placed in the correct directories.
Includes a clean and responsive Gradio UI for quick testing and demos.
Try it instantly without installing anything:
👉 KokoClone on Hugging Face Spaces
Recommended: Use conda for a clean environment.
git clone https://github.com/Ashish-Patnaik/kokoclone.git
cd kokoclone
conda create -n kokoclone python=3.12.12 -y
conda activate kokoclone
pip install torch torchaudio --index-url https://download.pytorch.org/whl/cpu
pip install -r requirements.txt
pip install -r requirements.txt
pip install kokoro-onnx[gpu]
KokoClone can be used in three ways:
Launch the Gradio app:
python app.py
Then open the browser interface to:
python cli.py --text "Hello from KokoClone" --lang en --ref reference.wav --out output.wav
from core.cloner import KokoClone
cloner = KokoClone()
cloner.generate(
text="This voice is cloned using KokoClone.",
lang="en",
reference_audio="reference.wav",
output_path="output.wav"
)
app.py → Gradio Web Interface
cli.py → Command-line tool
core/cloner.py → Core inference engine
inference.py → Example usage script
model/ → Downloaded TTS model weights
voice/ → Voice embeddings
This project builds upon:
Licensed under the Apache 2.0 License.