CLI Reference¶

RMVPE pitch estimator — ONNX Runtime inference

usage: rmvpe-onnx [-h] {download,predict} ...

Positional Arguments¶

command: Possible choices: download, predict

Sub-commands¶

download¶

Download the ONNX model

rmvpe-onnx download [-h] [--model MODEL]

Named Arguments¶

--model: Destination path for rmvpe.onnx (default: default model path)

predict¶

Run pitch prediction on an audio file

rmvpe-onnx predict [-h] [--model MODEL] [--device DEVICE] [--csv CSV]
                   [--confidence-threshold CONFIDENCE_THRESHOLD] [--plot]
                   audio

Positional Arguments¶

audio: Path to input audio file (any format supported by soundfile, typically WAV)

Named Arguments¶

--model

Path to rmvpe.onnx (default: default model path)

--device

Inference device to use: cpu, cuda, cuda:N, dml, or coreml (default: auto-detect)

--csv

Save results to a CSV file (columns: time, frequency, confidence); threshold is applied before saving

--confidence-threshold

Frames below this confidence threshold (0.0–1.0) are zeroed out (default: 0.03)

Default: 0.03

--plot

Show an interactive pitch plot after inference

Default: False