CLI Reference

Jabberjay ships a command-line interface installed as jabberjay.

Usage

jabberjay [-h] [-m MODEL] [-d DATASET] [-vis VISUALISATION] [-v] audio

Arguments

Argument	Description	Default
`audio`	Path to the audio file to analyse	—
`-m`, `--model`	Detection model to use	`VIT`
`-d`, `--dataset`	Training dataset for the VIT and AST models	`VoxCelebSpoof`
`-vis`, `--visualisation`	Spectrogram type for the VIT model	`ConstantQ`
`-v`, `--verbose`	Print model load and inference progress to stderr	off

Valid values

Models: AST, Classical, HuBERT, RawNet2, Spectra0, SpectraAASIST, SpectraAASIST3, VIT, Wav2Vec2, WavLM

Datasets: ASVspoof2019, ASVspoof5, VoxCelebSpoof

Visualisations: ConstantQ, MelSpectrogram, MFCC

Examples

# Default — VIT model, ConstantQ, VoxCelebSpoof
jabberjay interview.wav

# Choose a model
jabberjay interview.wav -m Spectra0
jabberjay interview.wav -m HuBERT
jabberjay interview.wav -m Wav2Vec2
jabberjay interview.wav -m RawNet2

# AST with a specific dataset
jabberjay interview.wav -m AST -d ASVspoof2019

# VIT with a specific visualisation and dataset
jabberjay interview.wav -m VIT -d ASVspoof5 -vis MFCC

# Enable verbose logging
jabberjay interview.wav -m WavLM -v

Output

Bonafide ✔️ (94.1% confidence, model=VIT)

or

Spoof ❌ (97.8% confidence, model=HuBERT)

Exit codes

Code	Meaning
`0`	Detection succeeded — the verdict is printed to stdout
`1`	Expected error (file not found, unreadable audio, invalid options) — a one-line `Error: …` message is printed to stderr
`2`	Invalid command-line arguments (unknown model/dataset/visualisation)