Guides and technical articles on AI music forensics, artifact removal, and production.
Closing the genre gap and the false-positive rate we promised to fix — a ground-up retrain on 27,476 tracks, a borderline-call review model, and the deployment bug we caught mid-rollout.
Negative results, evaluation leakage, and what we measured wrong — three failed experiments, a data-leakage discovery, and the real weakness hiding behind an illusory one.
Distillation and runtime — why light doesn't mean fast. Two parameter-count inversions, a causal real-time UNet, and a 50x speedup from a single ONNX surgery.
Listening cues, spectrogram analysis, and automated forensic tools for identifying AI-generated music from Suno, Udio, Stable Audio, and more. Includes benchmark comparison.
Step-by-step guide to removing the metallic, watery, swirly artifacts from AI-generated music. Covers RVQ ghosting, codec residue, and HF aliasing — what they are and how to fix them.
Chasing SONICS scores while real-world performance collapses — how a benchmark-saturated model went from SOTA to a 4.3% detection rate on the newest generator, and the fair-comparison harness we built in response.