What is ArtifactNet?
A lightweight, high-accuracy AI audio detection model
ArtifactNet is a 4.2M-parameter neural network that detects whether a music track was created by AI or a human. It works by analyzing spectral patterns that AI generators leave behind — invisible to listeners, but detectable by our model.
Unlike watermark-based approaches, ArtifactNet works on any audio file — no prior registration or cooperation from the AI generator is needed. It identifies content from Suno, Udio, Stable Audio, MusicGen, Riffusion, and emerging generators it has never seen before.
Each analysis returns a verdict (AI or Real), a probability score, and the suspected AI generator — processed in under 1 second per track.