DiffSingerMiniEngine
A minimum inference engine for DiffSinger MIDI-less mode.
Getting Started
- Install
onnxruntime
following the official guidance.
- Install other dependencies with
pip install PyYAML soundfile
.
- Download ONNX version of the NSF-HiFiGAN vocoder from here and unzip it into
assets/vocoder
directory.
- Download an ONNX rhythm predictor from here and put it into
assets/rhythmizer
directory.
- Put your ONNX acoustic models into
assets/acoustic
directory.
- Edit
configs/default.yaml
or create another config file according to your preference and local environment.
- Run server with
python server.py
or python server.py --config <YOUR_CONFIG>
.
API Specification
TBD
How to Obtain Acoustic Models
- Train with your own dataset or download pretrained checkpoints from here.
- Export PyTorch checkpoints to ONNX format. See instructions here.