Installation and configuration

Use Python 3.12, Git, uv and FFmpeg. Windows with PyTorch 2.10 CPU has been tested locally; real inference on other operating systems has not been verified.

git clone https://github.com/Sunwood-ai-labs/irodori-anime-voice-studio.git
cd irodori-anime-voice-studio
uv sync --locked --python 3.12
uv run app.py

Environment variables

Variable	Default / purpose
INFERENCE_DEVICE	cpu locally. The uv environment installs CPU wheels
CPU_THREADS	2; adjust to your CPU
GRADIO_SERVER_PORT	7860; change if already in use
HF_HOME	Hugging Face model cache location

$env:CPU_THREADS = '8'
$env:GRADIO_SERVER_PORT = '7863'
uv run app.py

After all required models are cached, set HF_HUB_OFFLINE=1 and TRANSFORMERS_OFFLINE=1 for offline use. Keep credentials outside source files. Setting cuda does not turn the CPU uv environment into a GPU installation.